How Will AI Transform the Future of Education?

TL;DR

AI tutors can personalize instruction, support mastery learning, and help students complete required academic work more efficiently. That shift could free teachers to become guides and motivators who foster curiosity, creativity, and human connection, while schools replace fact-heavy, just-in-case instruction with adaptable skills and just-in-time learning.

Transcript

• 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? (dramatic classical music) - So Steven, I've been a huge fan of your books for a long time.
• Thank you. - "Freakonomics" and "SuperFreakonomics," and they showed the power of data to upend conventional wisdom. And I, you know, now you'... Read More

Key Insights

• The central objective of education is to develop well-adjusted, curious people who are excited about the world. Levitt argues that centuries of layered changes have obscured this purpose, leaving schools focused on inherited processes rather than the outcomes students actually need.
• Just-in-time learning is better suited to an unpredictable technological future than just-in-case learning. Students cannot know which knowledge they will need decades later, so education should equip them to recognize gaps, find appropriate tools, and learn what a real problem requires.
• Fact and formula memorization is an inadequate response to a world where knowledge is readily accessible. Levitt believes students need the capacity to learn unfamiliar tools as circumstances change, especially because their work five or 20 years from now cannot be predicted reliably.
• Traditional grading incentives can distort teaching toward difficult, overcomplicated material that separates students into performance categories. Levitt acknowledges that educators may teach things that matter less because grades require them to distinguish students who receive A's from those who receive B's.
• Effective AI tutoring requires more than producing answers or agreeing with students. Gomes notes that AI systems can be sycophantic and affirm incorrect responses, so Google's LearnLM experiment is being tuned around characteristics associated with effective human tutoring.
• Mastery learning works by adapting instruction to what each student knows and does not know. Technology can let students skip familiar material, revisit concepts when necessary, and receive bespoke teaching at their current level instead of moving through one uniform sequence.
• AI tutors could make required academic learning substantially more efficient. Levitt's school is based on the expectation that material tested on the SAT or ACT can be taught in about one quarter of the usual time, leaving more hours for engaging educational experiences.
• Teachers remain fundamental because human relationships motivate learning. Levitt wants teachers to become cheerleaders and guides who connect with students, help them understand who they are becoming, and remove obstacles, while AI and other technology support personalized academic instruction.

Questions & Answers

Q: How could AI transform the future of education?

AI could transform education by giving students personalized tutoring at the level they currently understand. Such systems could identify what students know, skip material they have mastered, and provide repeated opportunities to learn difficult concepts. This could make required academic instruction more efficient and give teachers additional time to motivate students, build relationships, encourage creativity, and guide meaningful learning experiences.

Q: What is the difference between just-in-case and just-in-time learning?

Just-in-case learning teaches subjects, facts, and formulas because students might need them many years later. Just-in-time learning begins when a person encounters a real problem and then acquires the tools needed to solve it. Levitt argues that the second model fits a rapidly changing world because students cannot predict what work they will do five or 20 years from now.

Q: Why is memorizing facts and formulas insufficient for future learners?

Memorizing facts and formulas is insufficient because future work and technological demands are difficult to predict, while extensive knowledge is already available at people's fingertips. Students need to become capable of encountering unfamiliar problems, identifying what they do not know, and learning the necessary tools. Education should therefore develop adaptability and learning ability, rather than treating stored information as its primary outcome.

Q: How can AI support mastery learning in schools?

AI can support mastery learning by adapting lessons to each student's knowledge and progress. Students can skip concepts they already understand and revisit concepts they have not yet mastered, rather than advancing according to a uniform classroom schedule. Levitt describes this as bespoke teaching enabled by technology and makes it the central instructional idea behind his new high school model.

Q: What role should teachers have when students use AI tutors?

Teachers should remain responsible for motivation, personal connection, guidance, and helping students understand who they are becoming. Levitt describes teachers as cheerleaders and guides who remove barriers to student growth, while Gomes emphasizes that humans remain fundamental because people learn through other people. AI can handle more personalized academic practice, but it does not eliminate the importance of supportive human relationships.

Q: Why can traditional grading incentives weaken education?

Traditional grading requires educators to separate students into categories such as A and B performance. Levitt says the easiest way to create that separation can be teaching material that does not matter in difficult or overcomplicated ways. This incentive can pull classrooms away from curiosity, adaptability, and meaningful understanding because evaluation becomes intertwined with the design and delivery of instruction.

Q: What problems must developers solve for AI tutoring to work well?

Developers must ensure that AI behaves like an effective tutor rather than simply agreeing with the learner. Gomes warns that AI can be sycophantic and may tell students they are correct when they are not. Google's LearnLM experiment is therefore being tuned around known tutoring characteristics, while its developers work with teachers and school systems to understand what succeeds and what needs improvement.

Q: How could schools use time saved by personalized AI instruction?

Schools could use saved time for engaging activities that develop curiosity, creativity, discussion, and personal growth. Levitt expects his school to teach SAT or ACT material in about one quarter of the time currently used, but he identifies the freed time as the harder challenge. The goal is to create an exciting curriculum instead of using old instructional methods to warehouse students for the day.

Summary & Key Takeaways

• Steve Levitt argues that education has accumulated incremental changes while losing sight of its central purpose: developing well-adjusted, curious people who are excited about the world. Because future work and technology are unpredictable, students need the ability to acquire new tools when problems arise, rather than memorizing large collections of facts and formulas.

• AI tutoring could provide personalized instruction at each student's current level, skip material already mastered, and repeat material when necessary. Levitt's school is built around using technology for mastery learning, with the expectation that tested academic content can be taught in about one quarter of the time currently required, freeing much of the school day.

• Ben Gomes emphasizes that effective educational AI must behave like a tutor rather than merely provide agreeable responses. Humans remain essential for motivation and connection, so teachers could become cheerleaders and guides who remove obstacles, encourage creativity, and help students understand who they are becoming while technology handles more routine instruction.