AI extends the mind… and it can also amputate it.
We delegate writing, memory, critical thinking, creativity. GPS already eroded our spatial orientation; ChatGPT and Midjourney are eroding something wider: the habit of doing the mental work internally before outsourcing it.
Plato warned about writing 2400 years ago: it would “produce forgetfulness in the souls of those who learn it, because they will not practice their memory.” Today that warning is about prompts: when the shortcut becomes the norm, what atrophies is not the tool, but the muscle that no longer gets exercised.
McLuhan called it “desperate auto-amputation”: we extend a faculty outward and, to tolerate the new scale, we numb parts of ourselves. Heidegger spoke of Gestell, the enframing that renders everything as standing-reserve, as usable resource. The result? A generation that seems to know everything… while knowing almost nothing from the inside.
But it isn’t inevitable.
There is still time.
We can still resist.
1) How AI accelerates cognitive atrophy
Atrophy happens not because “AI is bad,” but because of a pattern: delegating before attempting. When the daily sequence becomes “ask → copy → paste,” the loop “read → try → fail → adjust → understand” stops repeating.
The cost is delayed. You notice it later: when you must write without help, argue without scaffolding, remember without search, decide without recommendations. The loss is quiet because it masquerades as productivity.
2) Extensions that shrink us (hot vs cool)
McLuhan distinguished “hot” and “cool” media by how much participation they demand from the receiver. However we classify specific tools, the practical point remains: some technologies invite participation, and others replace it.
GPS replaces an internal map. Generative models can replace an internal draft. The danger is not using extensions; it is using them in a way that systematically reduces participation.
When the world becomes overly assisted, the mind adapts to a low-friction climate. And as with the body: without resistance, there is no strength.
3) The ontological risk of delegating being
We do not only delegate tasks; we delegate criteria. We outsource what counts as “good,” “enough,” “interesting,” “true.” When that becomes habitual, life reorganizes itself around what is suggested.
Here the risk becomes ontological: not merely losing skills, but losing authorship—becoming an operator of interfaces, someone who selects from pre-built menus instead of someone who discovers, persists, endures the discomfort of not knowing, and through that process changes.
If everything shows up as resource—and we ourselves become a resource to optimize—the horizon of the human narrows without announcing itself.
4) A practical frame to recover mental effort
Resisting is not “never use AI.” It is reintroducing effort where we now default to outsourcing. A few simple, consistent practices can shift the trajectory:
- Attempt first: before asking, write 10–15 lines of your own (or an outline), then request feedback.
- Use AI as tutor, not substitute: ask for counterexamples, guiding questions, evaluation criteria; avoid “do it for me.”
- Assistance-free blocks: reserve windows (e.g., 45–90 minutes) for reading/writing/reasoning without models.
- Rebuild active memory: practice active recall (summarize without looking, teach out loud, reconstruct arguments).
- Occasional navigation without GPS: take familiar routes unguided to keep the internal map alive.
- Deliberate friction: choose projects that demand continuity, where progress isn’t instant.
None of this is heroic. It is hygiene.
The underlying question is simple: do we want to become increasingly efficient users, or increasingly capable beings?
We can still choose.