I keep coming back to a specific image. Not the one you’d expect: not the Terminator, not a glowing red eye, not servers humming in a bunker somewhere. The image I keep returning to is gentler and somehow more troubling: a doctor who no longer remembers how to diagnose without the algorithm.
It’s already happening. Quietly, in the margins.
Radiologists who trained alongside AI assistance are showing measurable degradation in the diagnostic skills their predecessors spent decades sharpening. The tool works so well that the human stops practicing the art. And when the tool fails, which tools always eventually do, there’s nobody left who remembers how to catch it.
This is not a story about robots taking jobs. That story is too simple, and frankly too old. This is a story about something more intimate: what we become when we hand the hard things over, and whether we like who’s left.
The gift you can’t return
Here is what I genuinely believe: AI is going to do extraordinary things for humanity. The optimists are not wrong. The vision of AI as a scientific collaborator, compressing decades of drug discovery, modelling protein folding, finding the climate solutions that human researchers keep almost reaching, is real and it matters enormously. If machine intelligence helps cure Alzheimer’s, or figures out room-temperature superconductors, or designs a fusion reactor that actually works, I will not be sitting here mourning the process by which it happened.
But.
There’s a version of progress that looks like gain on every spreadsheet and feels like loss in ways we don’t have spreadsheets for. And I think we’re entering it.
The philosopher Albert Borgmann wrote about what he called “device paradigm”, the way that modern technology hides its machinery and delivers results without requiring engagement. A hearth heats a room and demands something from you: you chop wood, you tend the fire, you gather around it. A central heating system delivers warmth and asks nothing. Borgmann’s point wasn’t that central heating is bad. It’s that the things we automate away often carried more than their functional payload. They carried practice, presence, and connection.
Multiply that insight by the most capable cognitive technology in human history and you start to feel the shape of what I’m worried about.
The labour question everyone keeps getting wrong
The economists will tell you, and they’re not lying, that technological unemployment historically resolves itself. The loom displaced handweavers; the tractor displaced farmhands; the spreadsheet displaced ledger-keepers. New jobs appeared. The standard of living rose. The story has a happy ending.
What the economists are slower to say is that the resolution never happens cleanly, never happens fast, and never happens equally. The handweaver who lost their livelihood in 1810 did not get a software engineering job in 1990. They lived through the in-between part, which was often brutal. The story has a happy ending on the hundred-year timescale and a much messier one on the human-lifespan timescale.
AI is accelerating the displacement curve while compressing the timeline. That’s new. Previous technological revolutions gave labour markets decades to adapt, long enough for the education system, for cultural norms, for social safety nets to partially catch up. What happens when the adaptation window shrinks from decades to years? What happens when the jobs that disappear first are not the low-skill jobs economists predicted, but the mid-skill cognitive jobs, paralegal, junior analyst, entry-level coder, that used to be the first rung of a professional ladder?
You can’t climb a ladder with no bottom rung. And right now, across knowledge industries, the bottom rungs are being removed.
The people who will feel this first and worst are not the workers at the top of the skill distribution, who will use AI as a force multiplier. They are the people who were on their way up – who needed the junior position to learn the craft, who needed the apprenticeship model to become experts. We are automating away the training ground.
What I keep wondering at 2am
I’ll be honest about where this gets personal for me, or at least where it gets philosophical in a way that keeps me awake.
I write. Not as performance, not for an audience specifically, but because the act of working a thought out in language teaches me something that thinking alone doesn’t. There is resistance in the sentence. You think you know what you mean until you try to say it, and then you discover that what you had was a feeling gesturing vaguely in the direction of a thought. The writing forces precision. The difficulty is the point.
What happens to that, to that particular friction, that particular form of cognitive development, when the option to have the machine draft it is always, frictionlessly available? Not “what happens to the essay as a product.” What happens to the essayist as a person?
I don’t have a clean answer. But I notice that we’re not really asking the question.
We debate AI in terms of outputs, is the essay good, is the code correct, is the diagnosis accurate. We are much quieter about what the process of generating outputs does to the human generating them. There is a growing body of evidence that reading on screens, versus paper, changes how we process information. That GPS navigation changes how we build spatial memory. That calculators, used early, change how we develop numerical intuition.
Cognition is not a fixed capacity we apply to tasks. It is something we develop through the tasks
themselves. And if AI removes the developmental friction from a wide enough range of
cognitive tasks, we may find ourselves, some decades from now, in possession of a very
powerful set of tools and a diminished capacity to know what to do with them.
This is not Luddism. I am not saying the tools are bad. I am saying that we have a consistent historical pattern of adopting tools without asking what the tool costs at the level of human development, and then being surprised by the answer.
The dependency problem is not hypothetical
There is a scenario that I think about with something approaching genuine dread. Not AGI gone rogue. Not the dramatic movie version. Something quieter.
Modern supply chains, power grids, financial systems, and communication networks are already deeply interdependent in ways that make them simultaneously efficient and brittle. A cyberattack on a Ukrainian power grid in 2015 caused blackouts for 230,000 people. A software update error in 2024 grounded thousands of flights globally. These are systems where the complexity has outrun any individual’s ability to understand the whole and where failure cascades in nonlinear ways.
Now imagine layering AI decision-making, deeply integrated and trusted, across healthcare systems, logistics networks, infrastructure management, and financial markets. The AI works brilliantly, until it doesn’t. And when it doesn’t, you are in a world where the humans who used to know how to run these systems manually retired ten years ago, where the institutional knowledge is gone, where the fallback is as degraded as the primary system.
This is not science fiction. It is systems engineering. Complex systems fail. The more we rely on AI to manage complexity, the more we must also invest in what happens when the AI is wrong, compromised, or simply encounters a situation outside its training distribution.
We are not investing in that. Not remotely proportionally to how fast we’re integrating.
The question underneath all the other questions
I want to name the thing I think is really at stake, underneath all the economic and technical and political arguments.
For all of recorded history, human beings have found meaning – real, felt, existential meaning –
through competence. Through the experience of doing something difficult and doing it well.
Through being needed by other people in ways that required genuine skill, judgment, and presence. The surgeon who saves a life. The teacher who reaches a struggling student. The engineer who solves a problem no one else could. The parent who reads their child well enough to know what the child actually needs.
These are not just jobs. They are the architecture of a meaningful life. And AI is not coming for the outputs of these roles, it is coming for the experience of performing them. For the feeling of being the one who figures it out.
I don’t know what replaces that. I’m not sure anyone does. And I think the people building these systems, many of whom are brilliant, well-intentioned, and genuinely motivated by a desire to help, are so focused on what the technology can do that they are not spending enough time with the question of what it does to us. To the texture of human experience. To what it feels like to be a person who is good at something that matters.
There is a version of the future where AI handles disease and poverty and climate change and we spend our lives in leisure and creativity and connection. Maybe. I’m not dismissing it.
But there is another version where we optimise our way into a profound purposelessness. Where the machines are full of grace and the humans are quietly, politely, redundant.
And I’m not sure we’re building the tools to tell the difference between those two futures before we arrive in one of them.
I don’t have a tidy conclusion. I’m not sure there is one.
-Written by Fida Wafiq
Want to go deeper?
- “Technology and the Character of Contemporary Life” by Albert Borgmann: Directly referenced in the article, this foundational philosophy book introduces the “device paradigm” and explores how modern conveniences strip away the meaningful effort and connection found in traditional practices.
- “The Glass Cage: Automation and Us” by Nicholas Carr: A highly relevant exploration of the hidden costs of automation, focusing on how granting software dominion over our cognitive tasks diminishes our skills, structural engagement, and tacit knowledge.
- “Shop Class as Soulcraft: An Inquiry Into the Value of Work” by Matthew B. Crawford: For those fascinated by the article’s conclusion on finding existential meaning through competence, this book offers a brilliant defense of the human experience of doing difficult things well.
