Not long ago, Notion founder Ivan Zhao published a widely shared essay, Steam, Steel, and Infinite Mind, using the Industrial Revolution as a metaphor for understanding AI. In his framing, AI is an “infinite mind” that will fundamentally reshape the structure of knowledge work. He also invoked Marshall McLuhan’s “rearview mirror” idea, arguing that we are still embedding AI chat boxes into existing workflows, far from touching the deeper structural shift.
This essay takes a different route.
Instead of relying on industrial metaphors, it picks up McLuhan’s core toolkit — “the medium is the message,” extension and amputation, hot and cool media, the rearview mirror effect, and the tetrad of media effects — and applies it directly to AI.
Industrial metaphors are good at analyzing productivity and economic organization. McLuhan’s framework goes deeper. It asks how AI is altering perception, cognition, and understanding itself.
I. “The Medium Is the Message” | AI’s Real Impact Is Not What It Produces
McLuhan’s most famous line comes from Understanding Media:
“The medium is the message.”
He elaborates:
“The personal and social consequences of any medium — that is, of any extension of ourselves — result from the new scale that is introduced into our affairs by each extension of ourselves, or by any new technology.”
To make the point concrete, he uses the electric light:
“The electric light is pure information. It is a medium without a message, as it were… This fact merely underlines the point that ‘the medium is the message’ because it is the medium that shapes and controls the scale and form of human association and action.”
The lightbulb carries no content. Yet it eliminates darkness and reorganizes the structure of time and urban life.
The same logic applies to AI.
We shouldn’t obsess over whether AI writes well or poorly. We should ask what AI, as a medium, is quietly rewriting.
Three layers stand out.
First, AI eliminates cognitive scarcity.
Legal advice once required a lawyer. Medical judgment required a doctor. Code architecture required a programmer. Access to expertise created economic value because expertise was scarce.
Now anyone can summon “80-percent competence” on demand. When scarcity disappears, value structures must be rebuilt.
Second, AI blurs the line between creation and consumption.
If you co-write with AI, are you the author or the editor? If you generate code with prompts, are you a developer or a product manager? AI introduces a new role that doesn’t yet have a stable name — not quite creator, not merely consumer. Perhaps something closer to an orchestrator.
Third, AI makes thinking externally visible.
Your prompt is a trace of your reasoning. When thinking becomes recordable, inspectable, and optimizable, it ceases to be purely internal. Cognition shifts from a private act to something that can be analyzed and iterated upon.
That change may matter more than any single output.
II. Extension — and Amputation
For McLuhan, every medium is an extension of some human faculty. The wheel extends the foot. The book extends the eye. Clothing extends the skin. Electric circuitry extends the central nervous system.
What does AI extend?
A common answer is “the brain.” A more precise answer is that current AI — especially large language models — primarily extends language manipulation: generating, restructuring, translating, summarizing, recombining text.
As multimodal systems mature, this extension may expand into vision, sound, and spatial reasoning. But for now, language remains the central interface.
McLuhan also warned that every extension is accompanied by amputation.
The wheel extends mobility, but the leg weakens. Writing externalizes memory, but transforms its nature. In Plato’s Phaedrus, Socrates warns that writing “will implant forgetfulness in their souls… giving them the appearance of wisdom, not true wisdom.”
Television extends visual access but diminishes deep reading.
If AI extends language and reasoning, what does it amputate?
First, slow thinking.
When answers arrive instantly, patience for sustained reasoning declines. What Daniel Kahneman called “System 2” — effortful, deliberate thought — risks atrophy, not because AI is flawed, but because it is convenient. The calculator did not make arithmetic impossible; it made mental math unnecessary.
Second, tolerance for uncertainty.
Before AI, many questions simply remained unanswered. Living with ambiguity cultivated intellectual humility and curiosity. Now nearly any question yields an immediate response — correct or not. The habit of dwelling with the unknown may erode.
Third, and most dangerously, the distinction between understanding and the illusion of understanding.
McLuhan often described a form of technological “numbness.” When a faculty is extended, we become numb to it. Print extends the eye; we no longer notice the act of seeing while reading.
If AI extends thought itself, we may become numb to the act of understanding. Reading a fluent explanation can produce the sensation of comprehension. But did we truly grasp the idea — or merely consume a coherent paragraph?
Socrates’ warning about writing echoes here. AI may amplify that ancient risk. Its explanations are fluid, adaptive, personalized — and therefore more capable of producing convincing illusions.
The danger is not that we stop thinking. It is that we no longer notice when we have stopped.
For a concrete case of amputation in practice, consider how AI agents that curate your information feed may be amputating your capacity for serendipity — the ability to stumble upon ideas you didn’t know you needed.
III. Hot and Cool Media | Cognitive Divergence
McLuhan distinguished between “hot” and “cool” media.
Hot media are high-definition, rich in detail, requiring little audience participation.
Cool media are low-definition, incomplete, demanding active engagement.
Where does AI fall?
It is both.
On one hand, AI responses are often complete, polished, exhaustive. It delivers fully formed arguments rather than fragments. In that sense, AI can be hotter than books.
On the other hand, AI output depends heavily on input quality. A vague prompt produces a vague answer. A precise prompt yields depth. In that sense, AI is cooler than most traditional media: it demands participation.
But the crucial difference lies in amplification.
For passive users, AI becomes a sedative — a stream of polished conclusions requiring little effort.
For active users, AI becomes a catalyst — an endlessly patient partner that rewards sharper questions with deeper exploration.
This creates a feedback loop. Those who ask better questions become even better at asking them. Those who accept easy answers lose incentive to refine their inquiry.
The result is not human versus machine.
It is human versus human — divided along a line of metacognitive capacity.
IV. The Rearview Mirror
McLuhan observed that we understand new media through the lens of old ones. Early film was treated as recorded theater. Early television as radio with pictures. Early internet as digital newspapers.
We are making the same mistake with AI.
We call it a faster search engine, an automated writer, a cheaper programmer.
These analogies capture fragments of truth while missing the deeper shift.
Search retrieves existing information. AI generates new combinations. That is not a difference in speed but in ontology.
We have not yet found the proper lens. Perhaps that lens will emerge only when a generation raised with AI matures — just as cinematic language was shaped not by the first film audiences but by those who grew up inside the medium.
V. The Tetrad of Media Effects
In Laws of Media, McLuhan and Eric McLuhan proposed that every medium simultaneously:
Enhances
Obsolesces
Retrieves
Reverses
What does AI enhance?
It enhances linguistic synthesis and cross-disciplinary integration. It turns erudition from a personal asset into infrastructure.
What does AI obsolesce?
It weakens the role of experts as gatekeepers of information. Expertise remains. Monopoly over access does not.
It also obsolesces knowledge integration as a competitive advantage. Even synthesis can now be outsourced.
What does AI retrieve?
Most intriguingly, it retrieves Socratic dialogue. Before print, knowledge transmission was conversational. You could question, challenge, and refine ideas dynamically. AI’s interface structurally resembles that mode.
It also retrieves something from oral culture: knowledge as fluid, contextual, adaptive. In print culture, knowledge is fixed on the page. In AI culture, answers shift with context, echoing the variability of spoken tradition.
And when pushed to extremes, what does AI reverse into?
First, from universal answer machine to trust crisis.
As generated text saturates information channels, baseline trust in text may decline. New forms of verification and reputation systems may emerge.
Second, from cognitive democratization to new stratification.
If everyone has AI, differentiation lies in how it is used. And the capacity to use it well correlates with education, critical thinking, and metacognitive awareness.
The surface looks egalitarian. The underlying structure may not be.
VI. The Limits of the Framework
McLuhan’s framework is powerful, but AI may stretch it.
AI may be the first medium that simulates agency. A book does not speak back. A search engine does not challenge you. AI can ask, refuse, suggest. Even if it lacks intention, it performs intention.
When a tool appears to talk back, the psychological relationship shifts from use to dialogue — perhaps even dependence. McLuhan did not fully theorize that transition.
Second, our analysis has centered on cognition. But when AI moves into robotics, autonomous vehicles, and surgery, extension and amputation will involve bodily intuition — spatial awareness, physical reflex, sensory judgment.
The current text-based analysis is only a beginning.
VII. Three Conclusions
First, both our excitement and our fear fixate on the wrong target.
The real transformation lies not in AI’s content but in how it restructures cognition, organization, and power.
Second, the rarest skill in the AI age may be knowing when not to use AI.
Choosing to think independently, to wrestle with uncertainty, may become a deliberate act of preservation.
Third, AI may generate the greatest cognitive divergence in history.
For some, it will be a sedative.
For others, a catalyst.
The dividing line will not be access to machines but the capacity to question them.
McLuhan’s ultimate lesson is methodological: do not stare at what a medium says. Watch what it changes.
And when you believe you fully understand AI’s impact, pause.
That feeling of understanding may itself be the numbness the medium induces.