What is the right way, if there is one at all, to integrate artificial intelligence (AI) technology into our education system? For Sal Khan, CEO of one of the world’s largest nonprofit education technology platforms, the answer is to take a step back and ask: Where can AI best complement current pedagogy? If a problem can be solved by pencil and paper, should we really be using AI instead? Khan joins Bethany and Luigi to discuss his recent book, “Brave New Words: How AI Will Revolutionize Education (And Why That’s a Good Thing),” in which he makes the case for why the education sector will not only survive but thrive in the age of AI. He shares his 17-year journey to build and grow his organization, which now provides over 10,000 videos on everything from integral calculus to art history, reaching more than 170 million registered users in over 20 languages, mostly for free. Together, the three talk about how and where AI can enhance the learning process: how AI has shifted Khan’s philosophy and approach to pedagogy, how it could democratize educational and economic opportunity, and what this all means for traditional modes of learning and instruction in schools and universities. They also discuss concerns about data ownership, Khan’s partnerships with tech companies, and the guardrails he proposes to protect education against the monetization of students’ data and the concentration of benefits to privileged children. Ultimately, he makes the case for why teachers aren’t going anywhere—and leaves aspiring nonprofit and civic leaders with advice on how to build a successful, mission-driven organization.
What is the right way, if there is one at all, to integrate artificial intelligence (AI) technology into our education system? For Sal Khan, CEO of one of the world’s largest nonprofit education technology platforms, the answer is to take a step back and ask: Where can AI best complement current pedagogy? If a problem can be solved by pencil and paper, should we really be using AI instead?
Khan joins Bethany and Luigi to discuss his recent book, “Brave New Words: How AI Will Revolutionize Education (And Why That’s a Good Thing),” in which he makes the case for why the education sector will not only survive but thrive in the age of AI. He shares his 17-year journey to build and grow his organization, which now provides over 10,000 videos on everything from integral calculus to art history, reaching more than 170 million registered users in over 20 languages, mostly for free. Together, the three talk about how and where AI can enhance the learning process: how AI has shifted Khan’s philosophy and approach to pedagogy, how it could democratize educational and economic opportunity, and what this all means for traditional modes of learning and instruction in schools and universities. They also discuss concerns about data ownership, Khan’s partnerships with tech companies, and the guardrails he proposes to protect education against the monetization of students’ data and the concentration of benefits to privileged children. Ultimately, he makes the case for why teachers aren’t going anywhere—and leaves aspiring nonprofit and civic leaders with advice on how to build a successful, mission-driven organization.
Read a review of Sal Khan’s book on ProMarket, written by Capitalisn’t team member Matt Lucky.
Sal Khan: Just know why you’re doing something. It shouldn’t be about AI. It should be about, what problem are you trying to solve? If the problem you’re trying to solve can be solved with pencil and paper, solve it with pencil and paper. But if it’s AI, use AI. People not just at schools but everywhere keep falling into this trap. When the iPad came out, Los Angeles spent some ridiculous amount—I’m tempted to say it was something like $500 million—on iPads, and then they just gathered dust because they didn’t know what they were going to use them for.
Bethany: I’m Bethany McLean.
Phil Donahue: Did you ever have a moment of doubt about capitalism and whether greed’s a good idea?
Luigi: And I’m Luigi Zingales.
Bernie Sanders: We have socialism for the very rich, rugged individualism for the poor.
Bethany: And this is Capitalisn’t, a podcast about what is working in capitalism.
Milton Friedman: First of all, tell me, is there some society you know that doesn’t run on greed?
Luigi: And, most importantly, what isn’t.
Warren Buffett: We ought to do better by the people that get left behind. I don’t think we should kill the capitalist system in the process.
Luigi: Of the many aspects of capitalism we discuss on our podcast, some are particularly dear to us. One is how AI will affect capitalism. The second is the role that not-for-profit institutions play in capitalism. The third is how we can offer equal opportunities in a world that is very unequal. And, finally, we’re very interested in the future of universities. Is the business model broken? What is the future? What will substitute for it?
Bethany: It’s rare to find a person who encompasses all four themes at once. One of these people is Sal Khan, the founder of the eponymous Khan Academy.
If you lived on Mars for the last 20 years, let me tell you that Khan Academy is a not-for-profit organization dedicated to providing a free, world-class education for anyone, anywhere.
Luigi, when did you first encounter Khan Academy?
Luigi: It was more than 10 years ago, when my daughter was in her last year of high school, or maybe before the last year of high school. I walked into her room, and she was watching YouTube, and I said, “What are you doing?” And she said, “I’m studying.” I said: “OK, give me a break. You’re watching YouTube. You’re not studying.” She said, “No, no, I’m actually studying.” And I said, “OK, show me.”
She showed me a video from Khan Academy that was showing how to do integrals. I was fascinated because I remember when I learned calculus—and my teachers were not great—I struggled at the beginning to understand some of those concepts, which are pretty abstract. I thought that the stuff was great. My second thought was, damn, I pay all this money to the Lab School, and this is the way you learn?
Bethany: Yeah, it’s so interesting. I had a less-impressive meeting with Khan Academy only because my older daughter really struggled with math for a period of time. I thought: “Oh, Khan Academy, I’ve heard of this thing. I’ll just plunk my”—I think, at that time, 7- or 8-year-old—“daughter in front of Khan Academy, and she’ll be able to use this and figure out all the math that she can’t do.”
In reality, for a kid who is really struggling with some learning issues, that’s actually not going to work for them. You need a human being to sit there with them and explain it to them. Despite the fact that I majored in math, my ability to teach my children math has always backfired in a really terrible way, for all sorts of reasons.
That is not to understate the incredible impact that Khan Academy has had. Its revenues in 2023 were around $100 million, but that figure has nothing to do with its impact because most of its products are given away for free. One hundred sixty-nine million registered users have spent a billion hours watching its videos. And these are not cat videos or songs—they’re math, science, history, and computer-science classes.
Luigi: I tried to do this back-of-the-envelope calculation. Say the average US public school spends roughly $17,000 per student per year and provides the students with 1,200 hours of education a year. If you do the algebra, it is $14 per hour.
If we had to price the contribution of Khan Academy at the same rate, its contribution to humankind is on the order of $14 billion at the cost last year of only $72 million. This is enormous value creation.
Bethany: That is a pretty incredible calculation. And Khan Academy is not limited to the United States. It has a platform in four languages: English, Spanish, Portuguese, and French. It is translated into more than 24 other languages.
Luigi: In 2014, Khan also created Khan Lab School, an independent school in Mountain View, California, associated with Khan Academy. More recently, in partnership with OpenAI, Khan Academy has developed Khanmigo, an AI-based tutor that can help you do system-personalized learning.
As he writes in his new book, Brave New Words, it is what Alexander the Great had with his teacher, Aristotle. If Alexander was having trouble with a concept—hard to imagine—Aristotle would slow down. If Alexander had a knack for understanding something—like, for example, military tactics—I’m sure Aristotle would have sped up his instruction or gone deeper. By having this one-on-one attention, the student never feels stuck or bored.
This isn’t just something that happened in the deep past. Today, top athletes and musicians, for instance, continue to learn through one-on-one coaching. However, without the support of teaching assistants or technology, it’s hard to imagine an individual teacher getting anywhere near this level of growth with a single coach and 30 students all at the same time.
Bethany: Of course, as your resident semi-cynical journalist, I have some skepticism about all of this. Just as I experienced with my daughter, if you plunk a child who is really struggling in front of the greatest computerized or AI teaching assistant in the world, is that really going to make a difference and address that child’s issues? OpenAI was looking for ways to make it seem harmless and helpful and, voilà, here’s education in the wonderful brand name of Khan Academy.
In his book, Khan argues that this will be a help to teachers, not a substitution, but really? Isn’t there a risk that in an era of budget cuts, this will further exacerbate the divides between rich and poor, as the rich get human teachers aided by AI and the poor kids get just AI?
Luigi: Bethany, thanks for being our local skeptic, but before we judge, let’s talk to Sal Khan.
You were a very successful hedge-fund manager, and at some point in your life, you decided to give up your career and start Khan Academy. Why?
Sal Khan: I’m flattered you say very successful. I was, I would say, a reasonably successful hedge-fund analyst. But as a lot of folks know—this was about 21 years ago—it just came out of conversation that I had with a cousin who was having trouble with math. I offered to tutor her. Then word spreads in my family that free tutoring is going on, and before I know it, there are a bunch of cousins, family friends, that I’m tutoring every day.
I just started writing software for them. A friend suggested that I make videos, too. When I started making those videos, my cousins infamously told me they liked me better on YouTube than in person. So, I just kept going there as well.
I was getting letters from people all over the world saying how it was helping them and sometimes in very significant ways. That’s why I did set up as a nonprofit in 2008, with a mission of free, world-class education for anyone, anywhere. In 2009, I quit my day job thinking that, hey, the world could benefit from this, and I was really enjoying it, and maybe we could make it work.
Luigi: When it came to educating your own children, you created a brick-and-mortar school, the Khan Lab School. Why?
Sal Khan: In 2011, some folks reached out to me and said, “Hey, you want to write a book?” I was initially skeptical. I was like: “Why would I write a book? I can communicate on YouTube. It’s an easy way to get stuff out there.”
But they said, “No, books have a certain gravity to them, et cetera.” I’m like, “OK, I’ll give it a shot.” I’m really glad I did because it does flesh out your thinking. The book was The One World Schoolhouse, and the first third of the book was about the history of education.
I said: “Oh, what was happening here is we had mass public education that came about 200, 300 years ago. We had to make compromises. We couldn’t give everyone a personalized education the way Alexander the Great had with Aristotle. We batched students together and moved them together to set pace.”
In the last third of the book, I allowed my imagination to go free a little bit and say: “Given what we know and the tools and the way the world is going, what should education look like? What would an ideal school look like?”
It wasn’t just about using technology. It’s about students being able to teach each other. It’s about mixed-age classrooms. It’s about mastery learning, which is just the idea that if you haven’t learned something well yet, you should be able to keep working on it.
It’s silly to get a C in Algebra 1 and then somehow try to get an A in Algebra 2, or it’s even sillier if you got a C in Algebra 1 and then you got an A in Algebra 2, that that C in Algebra 1 is still on your transcript. Clearly, if you got an A in Algebra 2, you know Algebra 1 as well. That should now be upgraded to an A.
It was also based on the idea that if you could learn some of your core skills more efficiently, you could have more time and space for some of the higher-order things: creating things, being entrepreneurial, going out into the world.
There were a few families, mainly Khan Academy employees, who were like, “Hey, let’s start a little pod school, homeschooling cooperative, around these ideas.” When the Khan Academy board caught wind, they were like, “Hey, it looks like you’re actually trying to start The One World Schoolhouse.” And I was like, “Yeah, pretty much.”
Some were skeptical that we should be starting a brick-and-mortar school. Some were like, “No, this is exactly what the world needs to push the envelope even more.” We did set it up as a separate entity from Khan Academy, so it isn’t the same organization, but that’s when we started Khan Lab School.
Now it’s a pre-K-12 school out here in Northern California. I’m learning a ton, I’ve learned a ton, but for the most part, I’ve been pretty excited about what it’s been able to do.
Bethany: What has it taught you about the balance between individualized learning and batch instruction? Are there places where batch instruction is called for, as much as that can seem like a mechanized version of learning? What has it taught you about that balance between being in person in a brick-and-mortar place versus being able to get the instruction you need online? Have your views on that changed?
Sal Khan: My views on that have only gotten stronger, but they haven’t changed. When I started this, I said, “Look, I don’t want my kids or anyone’s kids to only . . . No matter how well a tool like Khan Academy might be able to teach them . . .” Obviously, now, it’s in theory even better with things like AI.
That’s not what it’s all about. You don’t want children who don’t know how to interact, to deal with conflict, who don’t have support from other human beings, both their peers and faculty. If you look at the Common Core Standards or any state standards, they don’t explicitly call out things like this, like, “should be able to deal with conflict, should be able to be resilient, should be able to communicate their needs.” Actually, you’re able to do that more explicitly in an environment like this, when you can offload some—not all, some—of the core skills by leveraging technology.
But even when we’re leveraging technology, a lot of it is, OK, students are learning at their own time and pace, but now that you’re in the room together, interact with each other. Tutor each other or do something a little bit deeper.
Batch implies that you’re just moving everyone together at a set pace, regardless of whether it’s good for them or not. I’m not a fan of batch processing for education, but I am a fan of being in a cohort, and sometimes it is good for it to be an age-based cohort. You are going through certain life stages, for the most part, at the same time, certain levels of maturity.
And sometimes it’s really great to have mixed-age cohorts, which the traditional batch system usually misses. We’re seeing this at Khan Lab School. A lot of the bullying behavior, a lot of what I would call teenage angst behavior, the theory was that, look, they’re proto-adults, they’re capable of responsibility, but traditional schools don’t give them a lot of responsibility. We still treat them like kids.
But if you tell a 16-year-old, “Hey, you’re going to go mentor that 13-year-old,” it matures them, and they’re less likely to use bad language. They’re less likely to bully someone else because they feel like they’re a leader in the community. People are going to be listening to the things that they’re going to be saying. So, yeah, overall, it only reinforced how important it is.
I’ll say one other thing. I didn’t design it when we founded the school, but we’re hearing it loud and clear. We’ve had five graduating classes already, and they’re doing excellent in all the traditional . . . their standardized test scores, where they’re getting into college. I think the median student is getting into a college with a 10 percent acceptance rate. They’re doing all of that stuff great.
If you saw our types of test scores and placement, a school where kids are typically taking calculus, say, their freshmen or sophomore years in high school, you’d say: “Oh, that’s going to be a pressure-cooker school. That’s going to be a school where the kids are elbowing each other out.”
If you visit, our school’s the exact opposite of that. You would not see a more collaborative group of kids because they’re not going against each other. In a mastery-based system, they’re all working together so that they can all elevate and get to proficiency.
Luigi: I might be biased because I’m Italian, and I’m an alum of the Montessori school, but it sounds very much like the Khan Lab School is very similar to a Montessori school. One of the criticisms that people raised to Maria Montessori was that, at the end of the day, it was a school for the rich.
Certainly, it seems to me that Khan Lab School is a school for the rich. In Mountain View, I don’t want to know the tuition, but I don’t think it’s free as the rest of your program. What I find fascinating is the free part of your education because that really changes, in a way, the philosophy of education.
I was perplexed by some of the stuff you wrote because I don’t know whether you’re playing nice or you’re really nice or you’re tied to . . . When you are describing the history of education, it’s clear that when mass education was introduced, it was not a free gift that the upper class was giving to the lower class. It was training a workforce to go to work in a factory. All the discipline that is embedded in the traditional school is because you had to transform some anarchic farmers into disciplined factory workers.
Every mass education is, in some sense, distorted by the desire of the people giving this education. Every for-profit school is distorted by the desire to capture profit. You have found a mechanism, at least so far, to provide education to the masses basically for free. That enables you to have a completely different kind of education. If I were you, I would push this idea more of how your philosophy of education is completely different than the rest because it doesn’t need to provide a very high return.
Sal Khan: But we do want to make it have, arguably, a higher return. You’re absolutely right, Khan Lab School is not a free school. We have done it as a private school. We have another offering, Khan World School, which we’ve done with Arizona State, which is actually free to any student in Arizona. Anyone listening in Arizona should take a look at it. We have had to charge outside of Arizona because we don’t have a charter, but that’s a whole other thing.
One, I will defend the public-education system a little bit. Yes, the public-education system is engineered or designed based on a factory model because during the Industrial Age, that is how we learned to do things at reasonable quality at scale. I will say that it has done very positive things. Literacy rates have gone through the roof. The middle class has expanded dramatically. There’s a bunch of very, very good things that have happened.
But to your point, we now have tools where we don’t necessarily have to make that compromise, where mass education has to be one size fits all. Our main theory of action is, yes, there could be Khan Lab Schools and Khan World Schools, and we’re even exploring ways that people can independently work on Khan Academy and get credentials that can help them get into college or get a job.
But we think the real movement is going to happen by us working with traditional schools, meeting them where they are, because that’s where most students are, and helping those schools and those teachers personalize education that much more.
That’s where most of the energy of Khan Academy goes. You mentioned our marginal cost is close to free, but it does cost our budgets about $90 million a year, which we primarily fund through philanthropy. But most of that energy of that team, about 350 people now, is to build tools that we can go into schools and we can show those schools and train those schools, if you use it this way and it fits into your traditional curriculum, it’ll improve outcomes.
Now, I also do dream that one day we can revisit some of the things that we’re forcing kids to learn, but it’s important to meet the system where it is.
Luigi: Most of your area of intervention is K-12. What about university in general? Do you see that your next future is to create a Khan University in addition to Khan Academy?
Sal Khan: The simple answer is yes. There’s a young girl I just recently met. She’s a freshman at MIT, just finished her first year. It turns out she grew up in Kabul. The Taliban takes over, and her family flees to Pakistan. She has zero education. Khan Academy was her education, both when she was in Kabul and she couldn’t go to schools, and then when she was a refugee in Pakistan.
She gets in her mind that she wants to go to MIT. She applies. MIT’s super impressed by her, but they’re like, “We have no idea whether you’re ready for MIT or not.” We have a mechanism on schoolhouse.world where you can prove your knowledge. It records your face, your screen, while you take a Khan Academy assessment. You can validate your knowledge. UChicago is part of this as well.
She did that, and they admitted her. I met her when she came out to California as part of the Sam Altman AI bootcamp for the top engineers in the country. She was out here in Silicon Valley, and my wife and I had dinner with her.
That reinforces that we need to figure out ways of high school credentialing and then college credentialing. I hope in the next five to 10 years, we or other people are in partnership to find ways.
I don’t think it’s going to be a replacement for UChicago or a Harvard or Stanford. I think those are always going to be luxury goods. I hope we can create alternative pathways where they could be close to free but provide just as strong of a signal that you know your material. You’re just as qualified as someone from UChicago or Stanford or Harvard or wherever. So, yes, that is something I hope to do before I retire.
Bethany: Is there a key piece of insight that you got from having launched the Khan Lab School, from having this physical manifestation of your ideas, that led you to do Khanmigo any differently than you might have done it otherwise?
Sal Khan: Yeah. You mentioned Khanmigo. Khanmigo is our name for the AI assistant or the AI teaching assistant and tutor and other things, AI anything, that we’ve implemented in Khan Academy.
Yeah, there are some students, I would say maybe 5 percent, maybe 10 percent, who are self-motivated enough—and we can debate why they are self-motivated enough—to engage in something like that. We’re in a renaissance for those students. Those students can just go to town and learn at their own pace.
But most students need more support. They need accountability. They need human beings in their lives, too, and they need other things. Even those 5 percent, 10 percent, also can benefit from human-to-human interaction, et cetera.
That’s where I learned through Khan Lab School, and we’re learning every day working with school districts around the world, that the role of that teacher is essential. In this AI world, Khanmigo, a lot of people are like, “Oh, is this about replacing teachers?”
No, and I think it would be impossible to. I actually think as long as the teaching role evolves in the right way, it’s only going to become more and more valuable.
If you imagine teachers being people who only disseminate information, then there are other ways to disseminate information. It’s not just AI; it’s videos. Even when the first textbooks came out 150 years ago, teachers were afraid that it was going to take away their jobs, but now teachers can’t imagine teaching without a textbook. It’s all about the job evolving.
I think the best teachers today recognize that their job is not to just be the source of information. Their job is to make a really engaging, interactive environment when they’re in the classroom, to be able to personalize and differentiate as much as humanly possible, to be able to hold students accountable, to form human connections with them. If we can create tools—Khan Academy, Khanmigo—that can offload some work for the teacher, even in their prep work, to support students more in real time, it should free up the teacher to do more of that human-to-human work.
Luigi: I appreciate what you’re saying, but there is a risk. One of the roles of teachers, especially in the early years of education, is also to babysit the kids. That’s an important role. It’s not the most highly qualified part of being a teacher. The risk is that that role would remain, and the qualification would be taken by Khanmigo. It’s a bit like an Uberization of teaching where you reduce . . . With respect to taxis, the old-fashioned taxi drivers knew where to pick up the clients, what was the best route, all those characteristics that made them somewhat valuable. Today, the Uber driver is just a machine that pretty quickly will be replaced by a self-driving machine. Isn’t that risk in the process?
Sal Khan: There could be, but I don’t think that that’s how things are going to evolve. A lot of what we talk about is, ideally, you do have access to an amazing teacher, and we can raise the ceiling there, but if you don’t have access to a teacher or an amazing teacher, we can raise the floor.
On that latter case, in the extreme, yes, you can be in rural Alaska or a young girl in Afghanistan. You have no access to anything. Hopefully Khan Academy, Khanmigo, and we have another platform, schoolhouse.world, which is free tutoring and certification, can raise the floor for you.
We also know—and we’ve experienced all of this, sometimes, and this is happening, unfortunately, more and more in a lot of our schools—that you might have a substitute teacher for two or three months at a time, or you might have a teacher who’s not expert in that field.
Let’s say the school’s trying to teach financial literacy. There’s no financial-literacy expert in the classroom, so they get the PE coach. The PE coach might be a great PE coach, they’re not an expert in financial literacy, and now, all of a sudden, they’re teaching the financial-literacy class.
We hope that in those situations, once again, we can raise the ceiling and raise the floor, where at least you have something to lean on, where the students can continue to move forward.
But the ideal use case for the skilled teacher is always going to be that they can go that much further, especially if they’re supported with good tools to dive in, diagnose what’s going on with their students, do in-person interventions. There are teachers who are already doing this, but it takes heroic amounts of effort to personalize in a class of 30. Any of us, if we had to do that even an hour a day, we’d get tired before we start. There are teachers out there that are doing it for five, six, seven hours a day, and then they go home, and they spend 20 hours grading papers and this and that.
As you automate driving a car, yes, you just gave the example that already, out in my town here in Mountain View, at any intersection right now, I see five Waymos. It’s actually quite creepy. We’re now used to it now. It’s like these ghost taxis.
But I don’t think that that’s going to be the case in education because for my own children, or frankly anyone’s children, I want my kids to have that caring adult go up to them and say: “Hey, Sal, that was really great. I saw what you did there.” Or: “Hey, that wasn’t so cool how you just talked to Mary. How do you think you would you feel if she talked to you that way?” These things are going to be irreplaceable—and not just the teacher, I would say the teacher and the broader community that, ideally, kids are a part of.
Bethany: You have the best of intentions for Khanmigo and probably the wherewithal, from a personal standpoint, to make it happen. Do you worry more broadly that the era of AI in education could release something far less pleasant, in the sense that, in this day and age of budget cuts, too often, instead of a great teacher using AI, it will become a teaching assistant using AI or just AI for our least-privileged kids? What responsibility do you feel to stop that? Can you have any responsibility to stop that beyond putting out the best product that you can?
Sal Khan: I definitely have fears of AI and education, but I don’t have those fears.
I’ll start with those fears. I haven’t talked to one school district anywhere in the world, red state, blue state, anywhere, that even is hinting that they would think about using AI to replace teachers or use it as a justification for less-skilled teachers. If I did see that, I would advocate very strongly against it.
The places where AI is going to create a problem is cheating. As I write in my new book, Brave New Words, cheating isn’t a new phenomenon. It’s been there for a long time, but AI maybe has just made it a little bit more egalitarian. It’s a lot easier to cheat and it’s a lot harder to detect.
We are working on solutions there where the AI is working with the teacher, providing more transparency to the teacher on the process, on how something gets constructed. Instead of the student doing the paper in a vacuum, they do it with the AI, where the AI acts as an ethical coach, and then the AI can communicate to the teacher, making it transparent what they worked on.
If the student goes to ChatGPT or gets their sister to write the paper, our AI will tell the teacher: “I don’t know where this paragraph came from. We didn’t work on it. And it looks a little bit different than Jimmy’s other writing.” There used to be this whole class of stuff that we used to say: “Kids, go do that on your own. Do it at home.” They were already cheating, frankly.
Chegg, the public company whose market cap went from like $2 billion to zero when ChatGPT came out . . . The reality is Chegg was a cheating platform where, for $30 a month, kids could go put their homework up, and people would tell them the answers. Now, ChatGPT comes out, and it’s free to do the same thing. But it was already out there. I think it’s actually good that AI is putting a spotlight on this, so educators aren’t naive that kids weren’t already doing this.
Luigi: I agree 100 percent with that. This year I decided that with my students, I actually encouraged them to use ChatGPT because I think in the real world, you’re going to use it. Trying to prevent them from using it seems impossible.
I said: “You can use it. There’s one condition. I’m going to cold call you in class, and if you don’t know what you’ve written down there, you get a zero in class participation.” It seems to be working. So, I think that I’m very sympathetic to that view.
I’m also sympathetic to another thing that you write that I thought was very interesting, which is the fact that we learn in different ways. I learned it through my kids because of course, I learn my way, but then I saw that my kids were learning a completely different way than I learned, and that there’s not one way to learn. The great advantage, I think, of Khanmigo is that you can fine-tune it.
But let me play the devil’s advocate here. My concern is, now, who owns all the information? You are building a phenomenal dataset of how people learn, basically. You have individual characteristics that show how they progress, how they don’t, so that in the future, I can take a test and immediately be assigned to version A, or version B, or version C of Khan Academy.
What are the protections for this database? This is one thing I have to ask because, as we know, OpenAI is transforming itself into a for-profit, and when there is a lot of value at stake, it is very difficult not to be tempted. What prevents you from tomorrow becoming a for-profit that massively exploits all the data you have accumulated?
Sal Khan: Yeah, it’s an important question. Yes, I think OpenAI has completed its transformation into a for-profit. When I started Khan Academy, I didn’t know that the generative-AI revolution was going to come in 2025, but I did think about this notion of data . . . and even with old-school AI, just having people’s data, how they’re performing on different things, you could optimize, give them different treatments, et cetera, even back in 2008, 2009.
I did think that data is going to be very important, and that is one of the many reasons why I set up Khan Academy as a not-for-profit. Frankly, I didn’t want people to question why we are doing this.
We still take that very seriously, and it’s not just about the fact that we’re not-for-profit. We also are very clear with people that we will never commercialize the data. We will never sell it to anyone else. Whenever we use outside LLMs, they are not allowed to use the data.
Then there’s another aspect of data: I think there could be well-intentioned people who could say what I just said, but they haven’t passed their SOC 2 audits and all of that, and some hacker comes in and takes the data. That’s where I think it is important to have credible organizations that know how to protect the data as well, which I think we do quite well.
That’s where our DNA is. This is what we’ve been doing from day one. This is what we’ve promised the world. One of our advantages in this AI world—and I tell this to the team every day—is the trust people have in us. If you go back over the last 15, 16 years since Khan Academy became a real organization, there were a lot of people who showed up on . . . Some of them have even told me they were inspired by Khan Academy, and they said, “Oh, we’re going to democratize higher education, or we’re going to do this, we’re going to do that.”
They all had to sell out one way or the other. Not that they’re doing bad things, but they couldn’t stay true to that mission or vision. But hopefully, people see that we’re not just some fly-by-night thing that’s saying the right thing just to make people trust us inappropriately. We’ve been living by what we’ve been saying for 15, 20 years.
Bethany: Apart from the question about monetizing data, what about the question that data becomes a straitjacket for kids? You write about the potential for AI to be a companion going through your life and to help evaluate you as you roll into college applications. But does that then limit the potential of human beings to transcend themselves, to make a leap from one state of being to another mental state of being, because now there’s this thing that has been with them all along that says, “But that’s what you were, and thus that’s what you will be,” and that it becomes a limitation for certain types of intellectual leaps or leaps in being?
Sal Khan: I actually think that was already happening well before Khan Academy and AI. That example where you get a D in Algebra 1 in 9th grade, and then you become a serious student and get an A in Algebra 2 and precalculus and calculus, you’re still not going to be able to get into Stanford because you’ve got a D on your transcript, even though you turned everything around. Your transcript was already that data that can judge you pretty harshly at a time and later, when you get your act together, the world’s not willing to give you a second chance. It can even happen that you drop out of high school, and then later, in your thirties, you get serious, but now there’s a stigma associated with a GED or whatever else.
I think the solution here is around this notion of, I would say, mastery learning and lifelong learning. The mastery learning is this idea of, if you’re at a D level, you should always have the opportunity and the incentive to improve it.
On Khan Academy, that’s the way it works. If you keep getting something wrong, you can’t do it, but then eventually, you improve it, you can always improve what Khan Academy says, whether you know it or not. Then you couple that with lifelong learning because I think there’s another dynamic where we only have a window to really prove who we are. Then the rest of our lives, people judge us based on how we performed between the ages of roughly 14 and 22.
But if we have lifelong learning, where you could say, “Look, even at 30, I can master calculus as well as the best University of Chicago or MIT student,” then that should be all people know about you.
Bethany: If there’s one guardrail you could put around the introduction of AI in schools, one criterion you would put in place before a school could roll out AI throughout its K-12 education, what would it be? One prerequisite, maybe, would be a better word than a guardrail.
Sal Khan: Yeah, there’s a bunch I could list about transparency and security and all of that, but I would say the . . . This is not a guardrail; it’s just a guidance. Know why you’re doing something. It shouldn’t be about AI. It should be about, what problem are you trying to solve? If the problem you’re trying to solve can be solved with pencil and paper, solve it with pencil and paper. But if it’s AI, use AI. People not just at schools but everywhere keep falling into this trap. When the iPad came out, Los Angeles spent some ridiculous amount—I’m tempted to say it was something like $500 million—on iPads, and then they just gathered dust because they didn’t know what they were going to use them for.
Similarly, a million startups are claiming, “AI this, this, that.” But why are you using it? What outcome, what efficacy do you hope to push? Kick the tires, make sure that it’s real, make sure it’s not vaporware.
Luigi: In the book you describe how you were approached by Sam Altman to give a positive image to OpenAI and ChatGPT, and certainly you are contributing tremendously because Khanmigo is a great idea. Are you an investor in OpenAI?
Sal Khan: No. I wish I was, or I wish I bought some NVIDIA right when he called me or right when I saw GPT-4 six months before everyone else. But, no, I have a shockingly conservative investment portfolio. I live a very . . . I would say upper middle class. We have two cars in the garage, and I live in Silicon Valley in a nice cul-de-sac, but, no, it’s like a 2,500-square-foot house. So, no, I’m not rolling in it, but I’m very happy where I am.
Even to give Sam and Greg Brockman credit, I don’t think they were just looking for . . . I think, actually, most people in the field aren’t just trying to look good. I actually think they want to believe that what they’re doing can be good for humanity. I think they actually did call us because they thought we could do something that’s truly good, not just something that looks good.
Luigi: I’m the economist here, so I have to play a certain role. I want to ask you, you live in one of the most expensive, if not the most expensive, areas in the world, and you make a nice salary, but people who did what you did in Silicon Valley, they’re worth billions. I don’t know your value, but you’re not worth billions.
How do you feel about your choice of starting a not-for-profit now? What motivated you then, what motivates you now, and what are you telling other people who have to make this decision today?
Sal Khan: I’m always a glass-half-full person. My wife will tell you I’m good at always convincing, arguably deluding myself sometimes, that the grass is always greener on my side—which I think is a great way to live, by the way. I advise that.
But I don’t question for a second having Khan Academy as a not-for-profit. I have zero envy for anything other than what I have. I remember in the very early days of Khan Academy, Bill Gates, who has been a great supporter of ours, once told . . . I think he said it publicly in a newspaper article, “I envy Sal.”
When I first read that—this was the early days of Khan Academy—I was like: “What? Bill Gates is envying me?” Imagine that, OK, if you go from having $10,000 to $100,000, it is 10 times better. If you go from $100,000 to even a million dollars, maybe it’s 10 times better. But if you go from a million dollars to a billion dollars, it doesn’t necessarily . . . There’s diminishing returns in terms of your direct happiness.
I think what Bill was saying is that he really envied that I get to work on something all the time that is . . . One, he likes to learn for learning’s sake. I love to learn for learning’s sake . . . and really steep yourself in that and to be able to communicate that and feel like I get to be part of something that could hopefully empower hundreds of millions or billions of people. Yes, zero envy on my part. I almost feel like I’m showing off right now, so I’m going to stop talking.
Bethany: OK, Luigi, I may have started as a cynic, but I think I came out a believer. I can still see some downsides for how AI, particularly in schools, might be used to harm and not for good, but I think he’ll be standing in the way of that, not furthering it to the extent that that’s possible.
I have to admit, I’m really a sucker for this idea of being able to credential people in a different way and being able to use AI-based tools to change that D that you got in your freshman algebra class.
Luigi: Yeah. I actually have to say I am in love. Wikipedia and Khan Academy have been the greatest innovations of the last 30 years because they really have changed the world in a way that was unimaginable until very recently.
You’re much younger than me, but you do remember the time when people were selling encyclopedias, and they were paying $1,000, $2,000 for all these volumes, just to have at their fingertips something that now everybody around the world has at their fingertips. Everybody with access to a computer—which is not everybody, I have to say—has this.
Khan Academy is the same order of magnitude. I’m with you, it is not going to solve all the problems in education, but the availability of this to people who are highly motivated is phenomenal.
I’m not a child psychologist, but I think that Khanmigo also has the potential to help people who are less motivated. I was listening to his audio version of the book, and some of the ideas with which he motivates people with Khanmigo, which is basically a gamification of education, is incredibly powerful.
As I said in the discussion, I’m lucky because I learned in the way I was taught. I adapted, and the reason why I’m here is because the way I was taught fit the way I learn. But I was just lucky. A lot of people learn in a different way, and the ability to fine-tune this, I think, has enormous potential.
Bethany: I still do worry a lot. I think that these fundamental big questions we have about how AI is going to play out through our educational institutions and whether it really will be a force for good or whether there are places in which it can be a force for evil, I think those are big questions. But I at least have more insight into how it really could be a force for good. I see the offsets much more clearly than I did before we talked to him.
Luigi: My concern long term is that people will try to go around him or, God forbid, he passes away, what will happen? The story of OpenAI that starts as a not-for-profit and then changes is a potential threat.
The reason why I’m so adamant about the fact that the law should intervene to discipline the process is because otherwise, there will be abuse, is really important in this particular case because he’s going to accumulate a fortune. There are several not-for-profits that turned into for-profits and made a fortune.
Once I was told that, for example, there was a not-for-profit organization of farmers that was collecting the quality of every piece of soil in America. Then Monsanto came and said, “We want to buy you out.” And they said, “Oh, of course not, I’m a not-for-profit, blah-blah-blah.” “What about a billion?” “Yes.”
I think that when this amount of money is so outrageously big, it’s very difficult to resist. We need to reinforce the system so that it’s not that easy to go around it. I’m telling you, he’s going to sit on a gold pot, and a lot of people will try to get it, the gold pot, one way or another. From the way he speaks and what he has done so far, I trust him 100 percent. However, what about going around him or what about past him if he wasn’t around?
Bethany: But I just wanted to go back to one other answer he gave that I really, really liked because one of the criticisms of the ever-expanding role of technology in classrooms has been that it hasn’t done all that much. It hasn’t really improved outcomes.
I thought his answer that you have to know why you’re using the technology was so smart. I think that, more broadly, might be true of any of us when we interact with technology, too. If you want it to do it for you and to substitute your own thinking, then maybe that’s the line where you have to say, “This isn’t mine anymore.” But if you’re using it to augment your thinking or to help you learn something that you then know for yourself, then maybe that’s the way in which it’s OK. Maybe that’s true in the classroom as well as in life.
Luigi: Yeah, I certainly think that’s true, and I hope that we all start to use it. By the way, there’s no alternative because either we use it, or somebody will use it badly at our own expense. I think that there is no real alternative.
Bethany: Agreed.
Luigi: The only other thought is about the university. I was intrigued by the fact he thinks about the university because I think that it needs to be reformed. The cost is too high. If we want to educate the masses, we need to find a way to educate them at a cheaper rate.
Bethany: Yeah. It’s not just that, it’s also kids’ social well-being and psychological health, and the path he’s offering actually has a really optimistic possibility to it, because the more the world narrows to this very small list of schools that kids feel like they have to get into, and they can’t get that D in a class because that’s going to wreck their GPA, and they can’t experiment with learning something and then risk not doing well at it, the more narrow life becomes and the less adventurous and more stressed-out kids we are creating.
This idea that you can recredential yourself or that you don’t necessarily need the checkmark of a Princeton or a Harvard because you’ve credentialed yourself through something like Khan Academy and that you can make up for . . . You can try something that is a stretch for you, that your brain isn’t naturally inclined to and then learn it and master it and have that be your credential, it actually feels quite revolutionary, in a sense. It feels like a way off this terrible flywheel trap that kids are stuck in now.
Luigi: You’re absolutely right. You remember our episode on meritocracy with Adrian Wooldridge, when we were talking about the Chinese mandarins? The way they were learning was such a structural hierarchy of learning that it really impeded creativity and change. To some extent, the decline of China is the result of being stuck in that way. I feel that we’re going in that direction.
I think that Sal represents a very different path, a healthy one. The irony, or maybe it’s not the irony, is that I read that he’s attacked because he doesn’t have a degree in pedagogy. That’s, again, the credentialists saying you don’t have the right credentials, so you can’t do that.