Brave New Words: How AI Will Revolutionize Education (and Why That’s a Good Thing), by Salman Khan (Viking, 272 pp., $30.00)
Salman Khan has been one of the most prominent voices in education for some 15 years now, and deservedly so. Khan Academy, the platform he founded in 2008, has made education accessible to anyone with an Internet connection, producing free videos on math, science, history, and more, from the pre-K to the Advanced Placement level. It’s one of the best cases of technology genuinely supporting, rather than distracting from, children’s education.
Now, alert as ever to the impact of new technologies, Khan has turned his attention to artificial intelligence and its educational prospects. He got access to AI on the ground floor, meeting with OpenAI’s Sam Altman and Greg Brockman months before the public release of ChatGPT in 2022. Quick to see its potential, Khan developed Khanmigo, a GPT-4-powered “tutor for learners, sidekick for teachers.” As he argues in Brave New Words: How AI Will Revolutionize Education (and Why That’s a Good Thing), this is only the beginning of the benefits AI can provide for both students and teachers. But though the book is a helpful introduction to AI’s educational value, it leaves out a crucial step: promoting decentralizing policies that complement AI’s customizability for each parent and student.
Through several brief chapters, Khan considers a wide range of potential gains from AI. For students, the core benefit will be customization. Educators have long recognized one-on-one instruction as the best way to learn. “Bloom’s 2 sigma problem” refers to the finding by education psychologist Benjamin Bloom that students receiving one-on-one tutoring perform two standard deviations better than their classroom-educated peers. But it’s called a problem for a reason. It would be prohibitively expensive for every student to have a full-time, one-on-one tutor, so we settle for the classroom model, in which one teacher shepherds a herd of students at the same pace. This approach has obvious downsides: the average student does fine, but advanced students get bored, and underperforming ones fall behind.
Khan argues that AI could finally solve this problem. Working with a chatbot outside the classroom, each student can enjoy personalized assistance on the specific mathematical concept or science topic he’s struggling with, allowing for individual progress at the appropriate speed. And for disengaged students, the chatbot can design its lessons around the students’ particular interests, relating one student’s math practice to sports and another’s to music, for example. In this respect, AI would offer an improvement even over Khan Academy’s videos, which, however clear and informative, must speak to a general audience rather than an individual learner. AI can also provide personalized progress reports to parents, boosting student accountability and giving parents a better sense of what their children are learning.
Relatedly, AI will not only personalize education but will also expand the range of subjects available to a given student. Roughly half of American high schools, Khan notes, don’t offer calculus, 40 percent don’t teach physics, and more than a quarter don’t teach chemistry. Khan Academy has done great work to fill these gaps, but AI could do even better, thanks to the immense store of knowledge available online.
Expanded access to online information could indeed be a major benefit of AI education, akin to living near a good public library. But there is another fundamental component of education: the motivation to learn, whether out of love of a topic or because the teacher is forcing you to do your homework. In the best circumstances, great teachers instill in their students a love for a subject; at other times, they must settle for making their students practice their times tables. It is far from clear how well AI will be able to perform either of these tasks. If you hate reading, will you give ChatGPT the chance to change your mind, or simply close your laptop?
Khan recognizes this challenge. He finds that Khanmigo “is powerful for students who proactively seek out the AI’s help” but that “most students neither understand the capabilities of AI nor naturally ask for help.” Khan Academy’s videos are wonderful resources, too—for those who really want to learn the material. He says little about how AI can help those who simply don’t care. To be fair, this is a problem for traditional educational models, as well, as the sharp rise in absenteeism illustrates.
Teachers will also benefit from AI, Khan argues. He foresees AI taking over tedious administrative tasks like grading and designing lesson plans, freeing up teachers’ time and attention for actual instruction. AI could reinforce the “flipped classroom” approach, which leaves basic knowledge transmission to the students at home and reserves class time for discussion and personal instruction. These benefits seem plausible but are hardly revolutionary. For teachers, curriculum design is essentially a fixed, one-time cost; they may fine-tune their lesson plans over the years, but would AI really inspire them not only to change but also to improve those lessons, rather than repeating last year’s? Similarly, teachers could assign students AI-augmented lessons at home so that they would be ready for classroom discussion the next day—but if all students were to be equally prepared for such discussions, their individualized AI lessons would need to impart the same material, at the same speed, to everyone. So how “individualized” would they really be?
Khan insists that “there’s no job that is safer in the large-language-model world than teaching.” This assertion is hard to square with the claim that AI will revolutionize education. Insofar as the teacher’s job is to provide information, teachers may indeed be in trouble, because AI can do that. A teacher might reply that motivating students is also vital, and an AI teacher can’t do that. However, as witnessed by the growth of education savings accounts, micro schools, and homeschooling, many parents are finding that they can handle the motivational side of education themselves.
As far as education goes, then, the AI revolution couldn’t have come at a better time, as parents and policymakers find themselves increasingly open to new approaches and tools. For AI to realize Khan’s vision of personalized education, it will need to be complemented by similarly personalized parental-choice policies. AI’s benefits to education are unlikely to be realized as long as they remain confined to the Procrustean schoolhouse, which forces every student into the same approach with the same teacher. Coupled with policies promoting pluralism and choice, AI could indeed help facilitate a more customized and successful education system.
Khan’s thesis in Brave New Words—that AI will revolutionize education—thus needs both fleshing out and qualification. But if education reformers can extend their current win streak, they should be able to move things forward.
Photo: PhonlamaiPhoto / iStock / Getty Images Plus