If I show you the number 201, does it mean anything to you?
It could be the result of 5 × 40 + 1 or it could be the final line of a proof that took months and is only understood by five people on Earth. It could be a transcription error. As an answer, 201 tells you almost nothing. Without the question or the process that led to it, the output is informationally empty. This is not a new problem, but AI has made it easy to forget.
42 was never the joke people think it was
Douglas Adams explored this exact issue in The Hitchhiker’s Guide to the Galaxy, though it’s often remembered in a simplified form. In the story, a civilization builds the greatest computer ever created—Deep Thought—to calculate the answer to the ultimate question of life, the universe, and everything. After seven and a half million years of computation, Deep Thought announces it has finished.
The answer, it says, is forty-two.
The assembled philosophers and scientists are furious. This is clearly nonsense. Deep Thought patiently explains that the answer is perfectly correct; the real problem is that no one ever knew what the question was. Without the question, the answer cannot mean anything, no matter how accurate it is.
The joke is not that the answer is silly. The joke is that computation without understanding produces perfectly correct irrelevance. Deep Thought did not fail. The people using it did. (The movie was crap; read the books instead!)
Compression is not understanding
AI systems are extraordinarily good at compression.
They summarize.
They condense.
They abstract.
They produce answers quickly and fluently.
What they do not do is place you in contact with the causal structure of a subject. And that distinction matters more than most people realize.
Knowing the summaries of a hundred books about childcare does not make you even one percent more qualified to care for children than someone who has actually done it. The missing ingredient is not information, but contact with reality.
Imagine trying to summarize a novel like To Kill a Mockingbird or a poem like The Waste Land in a single paragraph—you might get the plot, the words, and even a few key images, but all of the depth, all of the subtlety, and all of the lived experience are gone. Now compare that to a student who uses AI to summarize lecture notes so they can regurgitate them in a final exam: the AI version might be accurate, it might even be coherent, but it still can’t substitute for the understanding that comes from thinking through the material yourself, wrestling with it, making mistakes, and discovering your own connections.
This is not a moral claim. It’s a practical one. Which would you prefer your medical practitioners to have done?
Where AI is misused
A common misuse of AI looks like this:
-
asking for conclusions instead of maps
-
asking for judgments instead of distinctions
-
asking what to think rather than what exists
In other words, using AI to replace the formative parts of thinking, rather than the clerical ones, produces something that feels efficient, informed, and authoritative but is structurally similar to memorizing answers without knowing the questions.
A better analogy than “effort”
People often frame this as a question of effort: “You still have to do the work." That’s true, but slightly misleading. A better way to think about it is nutrition. AI can provide you with fully pre-digested intellectual calories. They are fast, palatable, and convenient. Consumed occasionally and knowingly, they’re fine. Consumed as a primary diet, they lead to malnourishment. This is not because they are “bad,” but because digestion—the act of breaking things down yourself—is where nourishment actually occurs.
AI becomes intellectual junk food only when it is treated as a finished meal rather than as ingredients.
A good use of AI
A genuinely beneficial use of AI looks more like this:
-
asking it to surface concepts you weren’t aware of
-
asking for contrasting viewpoints
-
asking for bibliographies, not beliefs
-
asking it to scaffold exploration rather than conclude it
In this role, AI saves time without severing your connection to the subject matter. It accelerates entry into a field without pretending to substitute for experience or understanding. This is not anti-AI. It is aligned with how learning actually works.
Natural law still applies
No tool exempts us from reality. Effort avoided at the beginning tends to reappear later as confusion, fragility, or misplaced confidence. That isn’t a moral punishment; it’s simply how the world works. AI does not break this rule; it follows it.
The cost of understanding cannot be abolished; it can only be postponed. And postponed costs tend to arrive with interest.
The real mistake
The mistake is mistaking compression for comprehension, answers for understanding, and fluency for contact with reality.
If AI is treated as Deep Thought—a machine that produces answers in isolation from questions—the result will be forty-two all over again: correct, impressive, and useless. Used properly, however, it can do something genuinely valuable: help us ask better questions and recognize that the questions come first.
