Po-Shen Loh: Why Kids Must Learn to Grade, Not Just Do

Why Learning to Think Beats Learning to Execute

Po-Shen Loh isn't your typical AI doomsayer. As a math professor at Carnegie Mellon University and the former national coach of the US Math Olympiad team (2013-2023), he's spent decades watching brilliant students succeed — and more recently, watching AI systems outperform them. When Google's AI solved 4 out of 6 problems at the International Math Olympiad, problems specifically designed to be novel, Loh took notice. "The International Math Olympiad has six questions and all six are very, very original... the artificial intelligence was able to come up with solutions to four out of six, which is more than I can do."

On the homework problem: "Using AI to do your writing homework in school is like saying, 'I'm not going to run a mile for exercise. I'm going to drive my car one mile for exercise.' How much exercise you get? You get none." Loh argues that the fundamental skill being lost isn't writing itself — it's the logical reasoning that writing develops. The AI is called a Large Language Model for a reason. If kids outsource language to machines, they lose the substrate for clear thinking.

On the paradigm shift: "People used to go to school to learn how to do the homework and do the exams. Today, everyone needs to learn how to grade the homework." This framing cuts to the heart of what changes when AI can execute most tasks. The valuable skill isn't producing answers — it's evaluating them. Loh's entire educational philosophy now centers on this: teaching students to generate original approaches, then critically assess what they've created.

On why humans must team up: "Soon, you're going to have to work together to survive. The only way to get other people to want to team up with you is for you to authentically and deeply be a person who is motivated by creating value in the other. If you are not that way, you are a bad partner and people will not want to go and team with you." In a world where AI handles execution, human value comes from collaboration, empathy, and the ability to delight others. The "lone genius" archetype becomes obsolete.

On the test prep trap: Loh describes the "huge industry around test preparation and cramming" as actively harmful — not just inefficient, but counterproductive to developing the flexible thinking AI can't replicate. When students are drilled to recognize every possible problem type, they lose the chance to practice invention. "But even worse, it takes away the student's chance to invent."

On AI as a thinking amplifier: Loh himself uses AI extensively — but in a specific way. When visiting Nashville, he used AI to research the local music scene, not for a report, but to build his own mental model. "I wasn't using the AI to write the report for me. I was using AI to make myself better at that particular goal." This is the healthy relationship: AI as context provider, human as synthesizer.

5 Insights for Parents and Educators in the AI Age

• Mental fitness requires effort — Just as physical fitness requires physical exertion, cognitive abilities require cognitive exertion; outsourcing thinking to AI atrophies the mind
• The new skill is evaluation — Success shifts from being able to produce work to being able to assess work quality, catch errors, and identify when AI outputs are subtly wrong
• Deception immunity requires reasoning — People who can't think independently become vulnerable to manipulation; AI makes this worse by producing "very convincing, reasonable" content that may be biased or incomplete
• Empathy is a survival skill — In a world of AI automation, the only reason anyone would want you on their team is if you create authentic value for others
• Multiple perspectives are essential — Loh reads CNN and Fox News, tunes his social feeds to opposing viewpoints; with AI sounding authoritative on any topic, triangulating sources becomes critical thinking infrastructure

What This Means for Organizations Adopting AI

Loh's message isn't anti-AI — it's about using AI correctly. The workers who thrive won't be those who outsource all thinking to AI, but those who use AI to become better thinkers. This has implications for how companies should approach AI training: teaching employees to evaluate AI output, not just generate it. The competitive advantage isn't having AI tools — everyone will have those. It's having humans who can think independently enough to catch AI mistakes, synthesize novel ideas, and collaborate authentically with other humans. In Loh's framework, the organizations that win are those that cultivate "thoughtful" employees who view AI as cognitive infrastructure, not cognitive replacement.