Throughout human history, knowledge has been accumulating.
We learned how to grow and prepare food, build shelters, make tools, organize societies, and solve the basic problems of life. Then, generation after generation, we discovered new principles and invented new technologies on top of what previous generations already knew.
As our collective knowledge grew, specialization became necessary.
We developed doctors, lawyers, mechanical engineers, electrical engineers, software engineers, physicists, and countless other specialists. No individual could master everything, so we divided knowledge into disciplines and went deeper.
This model has worked extraordinarily well. And it still does.
But AI is beginning to change something fundamental: our accessibility to knowledge.
The problem I used to think about
When I was in high school, I sometimes had a strange thought.
If I had lived in Newton's time and wanted to study physics, perhaps reaching classical mechanics would have brought me close to the frontier of human knowledge.
But I live hundreds of years later.
If I want to approach the frontier today, classical mechanics is only the beginning. I have to continue through thermodynamics, electromagnetism, relativity, quantum mechanics, and increasingly specialized fields beyond them.
Knowledge doesn't just accumulate. The distance to the frontier grows.
I encountered the same phenomenon through mechanical engineering.
There was a time when mechanics itself could take you remarkably far. Then electricity became important, and mechanical systems incorporated motors, sensors, and control systems. Electronics became increasingly relevant.
Then came computers.
Now, if we want to build an automated machine, knowledge from mechanics alone is not enough. We may need mechanical engineering, electrical engineering, control theory, computer science, software engineering, data science—and increasingly AI.
Every new layer of knowledge creates possibilities for combining disciplines, while simultaneously making it harder for one human being to understand all of them deeply.
That is the paradox.
AI changes the cost of crossing boundaries
AI doesn't eliminate specialties. A doctor still needs medicine. An engineer still needs engineering judgment. A lawyer still needs to understand law.
We still have to start somewhere, and deep knowledge gives us the mental models required to reason well.
But the boundaries between specialties are becoming more permeable.
An engineer can explore a biological concept without first completing a biology degree. A doctor can investigate computational approaches. A mechanical engineer can prototype software, analyze data, or explore an unfamiliar mathematical method much faster than before.
AI dramatically reduces the cost of approaching unfamiliar knowledge.
And this matters because innovation often happens at the intersections.
If the cost of crossing from one field into another decreases, the number of possible combinations increases enormously.
That may accelerate the same process that has driven human progress for centuries: discovering something, combining it with existing knowledge, creating something new, and then using that creation as the foundation for the next discovery.
So what becomes scarce?
If knowledge becomes increasingly accessible, simply possessing information becomes less differentiating.
Something else becomes more important.
Knowing what questions to ask.
- What problem am I actually trying to solve?
- What knowledge might be relevant?
- What assumptions am I making?
- What am I missing?
- What could go wrong?
- What knowledge should I learn deeply today because I may need to make an important decision with it tomorrow?
These questions become especially important when AI can produce an answer almost instantly.
Because accessibility is not the same as understanding. And an answer is not the same as a correct answer.
The next skill is evaluation
Imagine facing an emergency five or ten years from now. An AI system gives you a recommendation in seconds. The important question will not be whether the AI can generate an answer. It probably can.
The question will be:
Can you determine whether that answer should be trusted?
That requires something more than access to knowledge.
It requires judgment.
We need enough understanding to recognize when an answer doesn't make sense. We need to understand assumptions and uncertainty. We need to know when a problem exceeds our own competence and when another expert—or another model—needs to be consulted.
In other words, as AI becomes more capable, evaluation may become as important as generation.
This is why I don't think AI makes learning less important.
I think it changes why we learn.
We are not learning only to store more knowledge in our heads. We are building the frameworks that allow us to ask better questions, recognize useful knowledge, evaluate answers, and apply them at the right moment.
Perhaps education should prepare us for questions we haven't encountered yet
For a long time, education has naturally focused on answering known questions.
But in a world where answers become cheap and accessible, perhaps a larger part of learning should be about discovering the questions worth asking.
- What problems will matter in ten years?
- What disciplines might unexpectedly intersect?
- Which fundamentals should I understand deeply enough that I can recognize when an AI-generated answer is wrong?
- And how do I build enough intellectual range that, when I suddenly need knowledge from a completely different field, I know how to approach it?
AI gives us unprecedented accessibility to accumulated human knowledge.
But accessibility alone is not wisdom.
The real opportunity may be learning how to navigate that knowledge—knowing where to look, what to ask, what to doubt, and when to act.
When answers become easier to access, the quality of our questions becomes more valuable.
And perhaps the most important thing we can learn today is how to ask the questions we will need tomorrow.