This essay draws on ideas from my book Attention Thief, available free at the end of this post.
The anxiety is everywhere. AI is making us dumber. We’re losing our ability to think, to write, to remember, to reason. Every time we let a machine do something our brains used to do, we surrender a little more of what makes us human. The concern is stated confidently, repeated often, and — I want to argue — built entirely on a misunderstanding of how cognition actually works.
Not a small misunderstanding. A foundational one. Two foundational ones, actually.
Argument One: Attention Has Always Worked This Way
You are breathing right now. You have been breathing your entire life — through sleep, through distraction, through grief, through joy, every minute of every day since you were born. And you have essentially no idea how you do it. You don’t consciously manage the rate. You don’t think about the diaphragm. You don’t monitor the oxygen levels. The whole extraordinary process runs without you, and you contribute nothing to it cognitively.
Is that a cognitive degradation? Did you lose something when breathing became automatic?
Obviously not. What happened is that a complex process got absorbed into the body’s predictive machinery, freeing attention for things that actually required it. That’s not loss. That’s the system working exactly as designed.
This is how attention has always worked. The brain is a prediction machine. Its deepest ambition — the thing it is constantly, relentlessly trying to do — is to convert the unpredictable into the predictable, the uncertain into the certain, the novel into the familiar. Every skill you’ve ever learned follows this arc: what once required intense conscious attention gradually becomes automatic, absorbed into the model, handled without effort. The novice driver grips the wheel with white knuckles and thinks about every gear change. The experienced driver holds a conversation while navigating rush hour. The attention that used to go to the mechanics of driving has been freed for other things.
We don’t call this cognitive degradation. We call it expertise.
AI is the same mechanism, operating at a different scale. When an AI system can handle a task that used to require human attention — pattern recognition, data summarization, routine correspondence, predictable analysis — human attention doesn’t disappear. It migrates. It moves, naturally and automatically, to whatever the AI hasn’t absorbed. To the edge of the model. To the things that are still unpredictable, still uncertain, still genuinely requiring the kind of engagement that only consciousness can provide.
This is not a choice we make. It’s not a policy we implement. It’s the way attention has always worked — moving away from the predicted and toward the surprising, away from what the model handles automatically and toward what the model cannot handle at all. AI doesn’t interrupt this process. It accelerates it. It pushes the frontier of human attention further out, into territory that is more complex, more genuinely uncertain, more fully demanding of what we actually are.
The person who worries that AI will make us stop thinking about pattern-based tasks is like someone worrying that literacy would make us stop memorizing. They were right. We did stop memorizing epic poems. We outsourced that to books. And what we did with the freed attention was — eventually — everything that followed. The loss of one cognitive burden is always the precondition for taking on a harder one.
Argument Two: You Never Owned Your Thoughts in the First Place
The deeper misunderstanding is about the nature of thinking itself.
When we say I thought of something, we mean something very specific. We mean: I manufactured it. I created it from nothing. The thought originated in me, was produced by me, belongs to me in the way that a table belongs to the carpenter who built it. Thinking, in this picture, is a creative act — a form of production, of fabrication, of bringing something into existence that wasn’t there before.
This picture is wrong. Not slightly wrong. Fundamentally, structurally wrong. And once you see what’s wrong with it, the whole AI-degrades-cognition argument collapses at its root.
Consider how your other senses work. When you look at a tree, your eyes don’t create the tree. The tree exists independently of your looking at it. Your visual system receives, processes, and relates to something that is already there — something that has its own structure, its own reality, its own existence prior to and independent of your perception of it. Vision is relational. It is an encounter between a perceiver and something perceived. The perceiver contributes the apparatus of perception. The perceived contributes the object. Neither alone produces the experience.
The same is true of hearing. When you listen to music, your ears don’t create the melody. The melody exists in the world — in the air, in the vibrations, in the structure of the composition. You relate to it. You encounter it. The music was there before you listened and will be there after you stop. Your hearing is the relationship between you and something genuinely external to you.
Thinking works the same way.
The thoughts you have — the ideas, the insights, the connections, the images that arise in consciousness — are not manufactured by you any more than the tree is manufactured by your eyes. They are patterns. They exist in the structure of language, in the accumulated record of human experience, in the conceptual space that is built up over centuries of people thinking about the same problems, encountering the same questions, arriving at the same nodes of meaning. When a thought “occurs” to you, you are perceiving a pattern that was already there. You are relating to something that exists in a space that is larger than you, older than you, not owned by you.
This is why thinking happens to us as much as we do it. Why the best ideas arrive unbidden, in the shower or at the edge of sleep. Why creativity feels, to those who practice it most deeply, less like manufacturing and more like discovery — like finding something that was waiting to be found, not building something from scratch. Why scientists speak of a theory being “in the air,” arriving independently in multiple minds at the same time. The pattern was there. Several people related to it simultaneously.
Now: if thinking is relational — if it is perception of patterns that exist in a shared space rather than production of thoughts that originate in a private mind — then what exactly is AI taking from us?
When AI relates to a pattern and produces an output, it is doing something structurally similar to what you do when you think. It is not reaching into your mind and removing something that belonged to you. It is encountering the same patterns you encounter, in the same shared space. The patterns remain. Your capacity to relate to them remains. The relationship between your mind and the structure of ideas is unchanged.
What changes is the company you keep in that space. You are no longer alone there. A new kind of perceiver has arrived, one that can relate to certain patterns faster and more comprehensively than you can. But the space itself is not diminished. The patterns are not used up. A thought thought by a machine does not become unavailable to you, any more than a tree seen by another person becomes invisible to you.
What AI Actually Changes
Put the two arguments together and a different picture emerges — not of cognitive loss but of cognitive reorientation.
AI is a pattern-prediction machine. It is, at its core, a system for absorbing the predictable — for taking vast amounts of human-generated content, finding the patterns within it, and producing outputs that reflect those patterns. It is extraordinarily good at this. Better than any human, at scale, for the kinds of patterns that can be extracted from large datasets.
But consciousness, as I’ve argued elsewhere, lives at the edge of the model — at the place where prediction fails, where the pattern runs out, where something genuinely new needs to be encountered. AI advances that edge. It absorbs more of the predictable, which means the unpredictable frontier — the place where human consciousness is most fully recruited, most fully alive — moves further out.
And thinking, understood as relational perception of patterns rather than private manufacture of thoughts, becomes not less important in an AI world but differently important. The human contribution is not the production of patterns — AI can do that. The human contribution is the quality of the relationship with patterns: the judgment about which patterns matter, the sensitivity to what is being missed, the capacity to encounter a pattern and ask whether it is true, whether it is good, whether it points toward something worth following. These are not computational operations. They are forms of wisdom, of taste, of the kind of engaged, uncertain, fully conscious encounter with ideas that no prediction machine can replicate, because prediction machines are defined precisely by their relationship to the already-known.
The worry about AI degrading cognition assumes that the valuable part of thinking is the part AI can do — the pattern-matching, the information retrieval, the production of plausible outputs. It ignores the part AI cannot do: the genuine encounter with the uncertain, the willingness to be surprised, the quality of attention brought to the edge of what is known.
That part has never been more important. And nothing about AI touches it.
These ideas are developed at length in my book Attention Thief — a full account of how attention works, why it gets stolen, and what remains for genuine human agency in a world of engineered capture. Free PDF and audiobook below.