How to be a creative thinker | Carnegie Mellon University Po-Shen Loh

YouTube URL: https://youtu.be/JpYA7WXkHyI
Host(s): Carnegie Mellon University
Guest(s): Po-Shen Loh

Podcast/Video Overview

Overall Summary

Po-Shen Loh argues that the AI age changes how we learn. It is no longer enough to “do the homework.” We must “grade the homework,” meaning we must critique, adapt, and improve AI-generated answers. His core message: teach people to think independently, not just repeat steps.

He shares a practical model for scaling creative thinking. Loh blends rigorous math, live human coaching, and improv-trained communication. He designs win-win-win systems that help learners, advanced students, and performing artists. The result is a program that teaches problem solving with delight, not grind. Viewers learn how to design learning that builds judgment, creativity, and real-world value.

References

• International Math Olympiad (IMO): A global high school math competition known for challenging problems.
• Combinatorics: A branch of math about counting, arrangements, and structures.
• Number theory: The study of integers, primes, and related properties.
• Improvisational comedy (Improv): Live, unscripted performance training that builds quick thinking and empathy.
• Intrapreneur: An employee who innovates and builds new things inside an organization.
• StackOverflow: A Q&A site where developers share solutions and code snippets.
• Business model: How a venture makes money to sustain its impact.
• Scalability: The ability to grow results faster than costs grow.
• Customer discovery: Talking to potential users to learn real needs before building solutions.
• “Grade the homework”: A metaphor for evaluating and improving AI or external outputs with judgment.

Key Topics Covered

From “doing homework” to “grading homework”

Under each topic, provide a 100 word summary of the discussion.
Po-Shen Loh explains a mindset shift. In the past, students practiced known steps to pass tests. Today, AI can attempt most tasks first. The human role is to check, refine, and decide. We must teach students to evaluate answers, not just produce them. That demands mental flexibility and clear thinking. He compares it to how engineers use StackOverflow. They search examples, then judge what fits. Learning now is about building judgment and creativity. This redefines success in school and work. It also changes what “studying” should look like.

• Main Point: The world now rewards people who can evaluate and adapt solutions.
• Core Argument: AI and open knowledge make production cheap; human value is expert judgment.
• Quotes:
• • “Everyone needs to learn how to grade the homework.”
  • “The first thing you will do is you will ask ChatGPT.”

Why novelty matters more than repetition

Loh criticizes the cram industry that drills unusual problems until nothing feels new. This steals the chance to invent. He recalls the 1980s, when novelty forced flexible thinking. He encourages using problems students have not seen. The goal is not high scores from memorized patterns. The goal is the skill to generate ideas under uncertainty. He wants learning to feel like exploration. That is how students develop confidence and creativity. Repetition has a place, but it should not be the main event. Practice should build adaptability.

• Main Point: Great learners practice facing surprise, not repeating scripts.
• Core Argument: Repetition reduces shock but kills creativity; novelty trains flexible minds.
• Quotes:
• • “It takes away the student's chance to invent.”
  • “The only way to do that is by giving them questions that they have never seen.”

A win–win–win education ecosystem

Loh describes an ecosystem joining three groups. Middle school learners get expert coaching. Highly skilled high school students become polished communicators. Drama and acting professionals gain flexible, paid roles as communication coaches. This system scales because each group benefits. Students get engaging live instruction. High schoolers gain leadership and real-world polish. Actors apply performance skills to education. This model also attracts families who want friendly, expert mentors. It builds a pipeline of thinkers who can connect and inspire.

• Main Point: Align everyone’s incentives to scale quality education.
• Core Argument: Systems grow when every stakeholder clearly wins.
• Quotes:
• • “Win, win, win. We have all three lined up.”
  • “Ideally that person you're talking to is friendly.”

The role of improv and communication

Loh took improv classes to communicate math more broadly. He learned that even “math nerds” can become engaging and clear. He then brought drama coaches into his program. Their job is to help high school coaches become friendly, confident communicators. This raises teaching quality and student joy. It also makes the experience feel alive, like a high-quality stream. Communication is not fluff. It is a core skill that unlocks understanding and trust. This choice increases both learning outcomes and scalability.

• Main Point: Communication training multiplies teaching impact.
• Core Argument: Improv builds empathy, clarity, and presence, which improves learning at scale.
• Quotes:
• • “Even a math nerd like me can take those classes.”
  • “Taught by math geniuses who are smiling.”

Focused curriculum that teaches thinking

The live program covers algebra, geometry, combinatorics, and number theory. These topics force reasoning and pattern discovery. They are chosen to teach thinking, not coverage. The problems come from middle school competitions. They are designed to be unfamiliar and interesting. Loh’s team uses them to train judgment and creativity, not memorization. The end goal is independence. Students should learn fast and then not need classes anymore. This is a bold contrast to programs that try to keep students enrolled forever.

• Main Point: Teach a core set of topics that develop general thinking skills.
• Core Argument: A focused, problem-driven curriculum creates independence, not dependency.
• Quotes:
• • “Our goal is to make it so that… you don't need any classes.”
  • “We cover these because these are a curriculum that teaches you how to think.”

Building a viable social enterprise

Loh believes money funds impact. His early “XP” site provided free explanations but lacked a business model. He iterated. First recorded classes generated revenue. Then he moved to live teaching with expert coaches. He insisted on friendly, high-quality human experiences. This version solved pain points for families and coaches. He estimates potential scale to 100,000 high school coaches and a million learners. The key is balanced incentives and sustainable revenue. The mission is social, but the model must work.

• Main Point: Sustainable impact needs a real business model.
• Core Argument: Revenue enables scale; aligned incentives protect quality.
• Quotes:
• • “Money is important for impact and influence.”
  • “We’re not doing a very good job of maximizing lifetime customer value… This is the social entrepreneurship.”

Customer discovery by showing up

Loh traveled city to city giving park talks. He carried AV gear and drew 50–100 people per event. This was customer discovery in the wild. He spoke with thousands of parents and students. He learned their needs and fears. Those insights shaped the live program and coaching model. He also goes into schools to teach sixth grade. He prefers direct involvement over assumptions. The core lesson: you learn the truth by doing the work with the people you serve.

• Main Point: Get out there and learn from real users.
• Core Argument: Direct, hands-on contact reveals real pain points faster than theory.
• Quotes:
• • “You cannot create value if you don’t interact with people.”
  • “I go and myself step in and start doing the work.”

Thriving in the AI “Wild West”

Loh frames the AI era as a Wild West with vast opportunity. The winners will be those who can create value. He urges everyone to be entrepreneurial or intrapreneurial. Build things that help others. Design systems where each party gains. Use AI for repetition. Use your mind for judgment, empathy, and strategy. Learn to research, test, and adapt. And remember: the goal of education is independence. Freedom to learn anything next.

• Main Point: The AI age rewards builders who create value and think critically.
• Core Argument: Pair AI tools with human judgment and entrepreneurial action.
• Quotes:
• • “Everyone should be a bit of an entrepreneur.”
  • “What the world needs now is a large scale way for everyone to learn how to grade homework.”

Key Themes and Insights

Teach judgment, not just procedure

Students must learn to evaluate AI outputs. Try treating ChatGPT’s response as a draft. Edit and test it. This builds critical thinking and autonomy.

• “Everyone needs to learn how to grade the homework.”
• “They need to be quick thinkers and analytical thinkers to understand whether… correct.”

Design win–win ecosystems

Create programs where students, mentors, and partners all benefit. This drives quality and scale without burnout.

• “Win, win, win. We have all three lined up.”
• “We will never have a high school student doing something unless I could explain… to their parent.”

Use novelty to train flexibility

Expose learners to unfamiliar problems. Reward exploration over rote. Creativity thrives under uncertainty.

• “The only way… is by giving them questions that they have never seen.”
• “It takes away the student's chance to invent.”

Communication is a core skill

Blend content expertise with improv-style coaching. Friendly experts beat unfriendly geniuses. Teaching is performance and empathy.

• “Ideally that person you're talking to is friendly.”
• “Even a math nerd like me can take those classes.”

Build sustainable social enterprises

Impact requires cash flow. Test business models early. Align incentives to protect learning quality.

• “Money is important for impact and influence.”
• “This is the social entrepreneurship.”

Practice real-world immersion

Leave the office. Teach in parks and schools. Observe, listen, and adapt. Insights live where the users are.

• “You cannot create value if you don't interact with people.”
• “I go and myself step in and start doing the work.”

Aim for learner independence

Design learning that ends dependency. The best program makes itself unnecessary over time.

• “Our goal is… you don't need any classes from anyone ever again.”

Actionable Advice and Takeaways

Immediate Actions You Can Take:

• When you use ChatGPT, critique it like a grader. Highlight errors. Ask for alternatives. Test results.
• Add one weekly session of unfamiliar problem solving. Use middle school competition problems.
• Take an improv or speaking class. Practice explaining tough ideas with warmth and clarity.
• Talk to five parents or students this week. Ask about their pain points. Take notes.

Long-term Strategies:

• Build an ecosystem with aligned incentives. Ensure every stakeholder clearly wins.
• Focus your curriculum on thinking-heavy topics. Use novelty to train flexible minds.
• Invest in coach development. Teach experts to be friendly, clear, and engaging.
• Tie revenue to impact, not seat time. Design for learner independence.

Questions for Reflection:

• What part of my work can AI do today? What judgment must I still provide?
• Where can I create a win–win–win structure in my product or team?
• How will I measure whether learners are becoming independent thinkers?

Noteworthy Observations and Unique Perspective

Overnight bus as research

• He seeks diverse human contact to understand needs in the real world.
• “You can't understand the real world unless you actually start going into various parts of the real world.”

Parent-centered accountability

• Every student task must be explainable to a parent as the best use of time.
• “Can we explain to their parent why… that thing is the best thing they can do?”

Make classes delightful, not dreary

• Production quality matters. Aim for live classes that feel like Twitch streams.
• “The class looks as good as a Twitch gaming stream, and it's taught by math geniuses who are smiling.”

Scale through mentors and performance skills

• Train high performers and pair them with drama coaches to multiply impact.
• “Lots of people with extraordinary drama skills… very interested in paid part time jobs to… coach.”

Success redefined

• Optimize to solve the customer’s problem fast, even if it reduces lifetime value.
• “We’re not doing a very good job of maximizing lifetime customer value… This is the social entrepreneurship.”

Companies, Tool and Entities Mentioned

• ChatGPT
• StackOverflow
• Carnegie Mellon University
• US International Math Olympiad (IMO) Team
• Broadway and Hollywood (as quality benchmarks)
• Twitch (as production benchmark)
• US middle school math competitions
• “XP” (free explanations website he built earlier)

Final Thoughts

Po-Shen Loh offers a blueprint for learning and leading in the AI era. Teach judgment, not repetition. Design ecosystems where everyone wins. Blend deep content with human warmth and performance. Get outside. Listen to real users. Build business models that fund meaningful impact. The result is independent thinkers who can learn anything next. Your next step: pick one area—curriculum focus, communication, or customer discovery—and make a concrete change this week.

Linkedin Ideas

From Doing to Grading: Learning in the AI Era

• Core Argument: Judgment beats repetition when AI drafts the first answer.
• Key Point: Teach students to evaluate, not just execute.
• Quotes: “Everyone needs to learn how to grade the homework.” “The first thing… you will ask ChatGPT.”

Build Win–Win–Win Learning Systems

• Core Argument: Aligned incentives unlock scale and quality in education.
• Key Point: Learners, mentors, and coaches should all gain.
• Quotes: “Win, win, win. We have all three lined up.”

Why Improv Belongs in STEM

• Core Argument: Communication is a multiplier for expertise.
• Key Point: Friendly, engaging teaching beats raw brilliance.
• Quotes: “Even a math nerd like me can take those classes.”

Design Learning for Independence, Not Dependency

• Core Argument: The best programs make themselves unnecessary.
• Key Point: Focus on thinking-heavy topics and novelty.
• Quotes: “Our goal is… you don't need any classes… ever again.”

Customer Discovery in the Park

• Core Argument: Show up where users are to find real pain points.
• Key Point: Direct contact beats assumptions.
• Quotes: “I started going to city after city… giving math talks in public parks.”

Blog Ideas

The New Homework: How to “Grade” AI and Think for Yourself

• Core Argument(s): AI drafts; humans judge. Independence is the new outcome.
• Key Themes to explore: Evaluation skills, research habits, testing.
• Key Point (s): Treat AI output as a rough draft. Iterate and verify.
• Quotes: “Everyone needs to learn how to grade the homework.”

Win–Win–Win: A Scalable Model for Creative Learning

• Core Argument(s): Alignment enables scale without losing quality.
• Key Themes to explore: Ecosystems, incentives, mentor pipelines.
• Key Point (s): Design programs where every stakeholder benefits clearly.
• Quotes: “Win, win, win. We have all three lined up.”

Why Improv Should Be Required for STEM Coaches

• Core Argument(s): Communication drives comprehension and trust.
• Key Themes to explore: Empathy, stage presence, clarity.
• Key Point (s): Pair subject experts with performance training.
• Quotes: “Ideally that person you're talking to is friendly.”

Four Math Topics That Teach You to Think About Anything

• Core Argument(s): Algebra, geometry, combinatorics, number theory build reasoning.
• Key Themes to explore: Novelty, problem selection, independence.
• Key Point (s): Use competition-style problems to train flexibility.
• Quotes: “We cover these because these are a curriculum that teaches you how to think.”

Customer Discovery on a Shoestring: Parks, Talks, and AV Gear

• Core Argument(s): Show up and learn from users in their world.
• Key Themes to explore: Guerrilla research, fast iteration, humility.
• Key Point (s): Direct contact reveals needs faster than surveys.
• Quotes: “You cannot create value if you don't interact with people.”