The Future of Education: Freakonomics’ Steve Levitt on AI and Personalized Learning AI tutoring

n a thought-provoking conversation, Steve Levitt, co-author of Freakonomics and founder of The Levitt Lab, and Ben Gomes, Chief Technologist for Learning at Google, explore the transformative potential of artificial intelligence (AI) in education.

Their discussion highlights the need to rethink traditional education systems, emphasizing personalized learning, the power of AI tutors, and the evolving role of human teachers in fostering curiosity and engagement.

Below, we delve into their insights and outline how AI can reshape education to better prepare students for a rapidly changing world.

The Need to Upend Conventional Wisdom in Education

Levitt begins by critiquing the outdated structure of education, which has evolved incrementally over centuries without addressing its core objectives. He argues that the system has lost sight of its goal to produce "well-adjusted and curious" individuals excited about the world.

“We've layered little changes on top of what already existed. And I feel like we've just so lost sight of the objective, which is to create a new generation of people who are well-adjusted and curious and excited about the world,” Levitt says.

This sets the stage for a broader discussion about how AI can address these shortcomings by enabling more adaptive and individualized learning experiences.

Gomes complements this view by pointing out that the rapid pace of technological change demands a shift in educational priorities.

Unlike the past, where students learned skills to last a lifetime, today’s learners must be equipped to adapt continuously.

“Technology is changing much more rapidly, so there's a need to adapt to technology,” Gomes notes, underscoring the importance of fostering adaptability over rote memorization.

Just-in-Time Learning vs. Just-in-Case Learning

A key theme in the conversation is the contrast between “just-in-case” and “just-in-time” learning.

Levitt criticizes the current system for teaching subjects like calculus or Shakespeare “just in case” they might be useful decades later, rather than focusing on skills relevant to immediate needs.

“I'm stealing it from one of my podcast guests… he talked about just-in-case learning, which is what we do now… Versus just-in-time learning, which is something which actually makes sense,” Levitt explains.

He argues that in a world where knowledge is readily accessible, cramming facts into students’ heads is less valuable than teaching them how to learn and apply knowledge dynamically.

Gomes agrees, noting that AI can facilitate this shift by providing tools that deliver knowledge precisely when it’s needed.

This approach aligns with how professionals like Levitt and Gomes themselves learn—by seeking out tools and information to solve problems as they arise.

“In a world in which you do not know what you'll be doing five years from now… cramming your head full of facts and formulas is not the answer to managing the future,” Levitt asserts, highlighting the need for education to prioritize adaptability and problem-solving.

The Promise of AI Tutors

Both Levitt and Gomes see AI as a game-changer in delivering personalized education, particularly through AI tutors. Levitt is particularly enthusiastic about their potential, noting that they are already as effective as “run-of-the-mill” human tutors and will only improve.

“If I ran the California Department of Education, my thought would be, ‘How do I get an AI tutor in every kid's ear?’… How is there not a stampede to get kids equipped with those tools?” he asks passionately, emphasizing the scalability and affordability of AI tutors compared to human tutors.

Gomes shares details about Google’s LearnLM project, which aims to create AI tutors that mimic the best qualities of human tutoring while addressing pitfalls like sycophantic responses.

“We've got this experiment called LearnLM, which is trying to solve the basic problems of tutoring… We have been tuning this LLM to actually be better about those characteristics that we know are true of tutors,” he explains.

By refining AI to provide constructive feedback and adapt to individual student needs, LearnLM seeks to make tutoring universally accessible and effective.

Levitt’s vision for The Levitt Lab further illustrates this potential. His school leverages technology to deliver bespoke education, allowing students to skip material they’ve already mastered and focus on areas where they need support.

“The crux of our whole school is that, by using technology, bespoke… teaching to kids at the level they are and being able to react to what they know and don’t know… That is such a better way of teaching kids,” he says.

This approach, enabled by AI, could reduce the time needed to master standardized test content by up to 75%, freeing up time for more meaningful learning experiences.

The Human Element in AI-Driven Education

While AI holds immense promise, both Levitt and Gomes emphasize the irreplaceable role of human teachers in inspiring and motivating students.

Gomes reflects on the importance of human connection in learning, citing his own experience with an inspiring history teacher who made the subject come alive through engaging discussions.

“I had a history teacher in school, very inspirational… He would set up history as like, ‘Here's a situation. Here are the parties.

What do you think is gonna happen?’… The discussion kept us engaged in a way that I'd never been engaged with history before,” he recalls. This underscores the need for AI to complement, not replace, human educators.

Levitt envisions teachers as “cheerleaders and guides” who help students discover their identities and overcome obstacles.

“In our school, what we're trying to do is take teachers and turn them into cheerleaders and guides, people who are helping the kids understand who they are and who they become,” he says.

By relieving teachers of rote tasks, AI can free them to focus on fostering creativity and personal growth, aligning with the original motivations that draw many to the teaching profession.

Gomes echoes this sentiment, noting that education is undergoing an “era of evolution” as institutions adapt to new technologies.

“Humans are still pretty fundamental. Like humans learn because of humans,” he says, emphasizing the collaborative role of teachers, students, and AI in creating effective learning environments.

Balancing Engagement and Efficiency

The conversation also touches on the dual components of education: the “want to” and the “have to.” Gomes points out that platforms like YouTube and Google Search inspire learning because users engage with them voluntarily.

“People come to those tools because they want to, not because they have to… Hopefully, the want to can become larger in this process using some of these tools of AI. And the have to can become more efficient,” he says.

By making mandatory learning more efficient through AI, educators can create space for exploratory, passion-driven learning that keeps students engaged.

Levitt agrees, expressing frustration with the current system’s reliance on rigid standards that stifle creativity.

“We would benefit a lot from a system in which teachers felt free to be more themselves and didn’t feel this pressure… to bomb through these four topics today because the state standards have 472 topics,” he says.

By reducing the burden of standardized curricula, AI can empower teachers to foster deeper connections with students and encourage creative thinking.

A Vision for the Future

Levitt and Gomes conclude with a shared optimism about AI’s role in shaping the future of education. Levitt reflects on his own teaching, admitting that his traditional classroom methods feel outdated, but his work with The Levitt Lab feels like “being part of the future.”

“In my own classroom, I'm not excited at all… But here, trying to create this, it really feels like being part of the future,” he says. Gomes shares this hope, looking forward to a time when their efforts yield meaningful change.

“I hope we have a conversation two, five, 10 years from now… that we spent our time well and it was fun, and we created amazing things for the world,” he says.

‍