Before AI, retrieving knowledge was part of learning.
Imagine I knew nothing about agriculture and wanted to learn about it.
I might go to a library, find the agriculture section, stand in front of the shelves, and look around. Because I had little background in the subject, I would probably start with a book called something like Introduction to Agriculture.
But while looking for that book, I would see many others.
Plant diseases. Soil science. Irrigation. Agricultural machinery. Crop production.
I came to the library with one broad question—I want to understand agriculture—but the process of searching would already begin to teach me what agriculture contains.
Retrieval itself becomes part of the learning journey.
It gives me opportunities to discover things I did not know existed—and therefore questions I did not yet know how to ask.
Now consider a subject where I already have some background.
I studied mechanical engineering, so if I wanted to understand the gears used in a car differential, I could approach the library differently.
I might first go to the automotive engineering section and find a book explaining how a differential works. That would help me understand why a car needs this mechanism.
Then I might go to the machine design or gear section and find a book specifically about gears. There, I could learn about geometry, dimensions, design principles, and perhaps the history of how these mechanisms developed.
The two paths give me different perspectives.
One teaches me the application:
Why does the car need this mechanism?
The other teaches me the principle:
Why is the gear designed this way?
And as my understanding develops, so do my questions.
I might begin with:
What is a differential?
But that can lead to:
- Why does a car need one?
- Why are these particular gears used?
- What problem were engineers originally trying to solve?
- What would happen if we designed it differently?
- Under what conditions would this design fail?
- What are the trade-offs?
That last group of questions matters.
Because good questions do more than help us understand something. They help us discover what could go wrong before it goes wrong.
What changes when we learn with AI?
Today, I can skip much of the retrieval process.
I can simply ask AI:
“How do the gears in a car differential work?”
Within seconds, I can receive a clear and structured explanation.
This is incredibly powerful.
But it also creates an interesting challenge:
If I receive a good answer immediately, will I continue asking questions?
The journey between question and answer has become dramatically shorter.
And that journey used to expose us to context.
While searching, we encountered neighboring subjects. We discovered terminology we didn't know. We saw different ways knowledge could be organized. Sometimes we even discovered that our original question was not the question we should have been asking.
AI can certainly help us do all of these things.
But there is an important difference:
We often have to ask for them.
And this matters not only for how much we learn, but also for the decisions we make.
Imagine asking AI:
“How should I design this?”
It might give me a very good answer.
But perhaps the more important questions are:
- “What problem am I actually trying to solve?”
- “What assumptions am I making?”
- “What could fail?”
- “What information am I missing?”
- “What are the consequences if my assumptions are wrong?”
- “What should I understand before making this decision?”
A good answer to the wrong question can still lead us in the wrong direction.
That is why I think the ability to ask the right questions becomes even more important when answers become easier to obtain.
The quality of our questions affects not only what we know, but also what we notice, what we anticipate, and what mistakes we may be able to avoid.
So how do we improve the quality of our questions?
I think it begins with two things.
First, curiosity.
A good answer should not always be the end of our thinking. Sometimes it should be the beginning of the next question.
- Why?
- What am I missing?
- What assumptions are behind this answer?
- What could go wrong?
- What would someone with a different background ask?
- Is there another way to look at this problem?
Second, clarity of purpose.
- What am I actually trying to understand?
- What problem am I really trying to solve?
- What decision will this information help me make?
- And what could happen if I misunderstand the problem?
The clearer we are about our intention, the better our questions can become.
For a long time, one of the challenges of learning was finding answers.
AI is making answers abundant.
But abundant answers do not automatically create better understanding or better decisions.
Perhaps one of the most important skills in the age of AI is therefore not simply knowing how to get an answer.
It is knowing how to question the answer, question the problem, and question our own assumptions.
Because the right question can help us learn something we didn't know.
It can reveal something we didn't know we needed to know.
And sometimes, it can help us recognize a mistake before we make it.
So perhaps the question we should keep asking ourselves is:
If AI can give me almost any answer, am I learning to ask the questions that will help me understand what is right—and recognize what could go wrong?