Published 2026-03-02
Tags: #education
Suppose we have come to a point where these AI tools can do anything you want. Is it still useful for humans to go through the hard work of educating themselves? Or here’s some food for thought: Q: Would you want a first-grader with access to an all-powerful-genie to be a nation’s president?
Hopefully not :) Here’s a counter claim: Education is necessary for one to want better. This is because to want better, you need to
Both of these require that you, not the AI, have the right reasoning and conceptual building blocks readily computable in your brain. Let’s go through each one of these points.
Real-life problems don’t have one optimal solution. They are filled with nuanced tradeoffs, and being able to understand these tradeoffs requires you to understand mechanisms across a wide variety of systems. As an example, if you are building an AI-backed product, you might want to reason: “How should we charge users for the inference cost? Should we show the user ads, which charge for their attention, instead of money, since more people have attention than money?”
A first-grader would have a hard time even saying this sentence out loud. For them to seriously deliberate on this tradeoff, they need to at least grasp concepts like attention, money, ads, inference, etc.), have the reasoning skills to map these options against their personal moral values that they hopefully have crafted to be mostly good for humanity.
Communicating with other humans with different wants is how you increase the chances that your want is a solution that benefits others — and usually the development of your values requires you to think about humanity at large, not just yourself anyway.
More pressingly, to communicate with other humans, you need to be able to do the above in REAL TIME! It’s not okay to me that while I’m discussing with my cofounder, they would say: “wait sorry I don’t understand the tradeoff between latency etc, give me an hour to catch up”. Negotiation, empathy, and collaboration happen at the speed of human conversation. This then requires that you, not AI, also have the mental agility to reason about these concepts in real time. Developing this mental agility requires hard work.
Before handing your 10-year-old the lamp (with a clawdbot sticker on it), make sure they know what to wish for and how to explain that wish to the rest of the world.
CS education? How deep down the mechanistic level does one need to shape one’s want? What kind of wants would require one to go through CS undergrad education and learn how to read Python code? See this post.
Another thought: Learning to shape your values. To blindly accept AI-generated work before thinking carefully about your want is to blindly accept an alien value.