SAL KHAN:
I think in 15 years, you literally will have every child—not child, every learner on the planet—has access to some form of what we could call “world-class education.” If you asked me even three, four years ago, and you said, “How are you gonna give every child rich, Socratic dialogue? How are you gonna give every child the ability to role play in a simulation, or to explore ancient Rome?” I was like, “Yeah, I’m not sure exactly how to do that right now, but Khan Academy’s going to be around well beyond me, so one day, you’re going to have AI, and you’re going to be in an immersive world, and you’re gonna have a link straight into your brain. You’re gonna have these lucid dreams where you’re in Rome, whatever.” That’s likely to happen in the next 15 years, and it’s likely to be very accessible.
REID:
Hi, I’m Reid Hoffman.
ARIA:
And I’m Aria Finger.
REID:
We want to know what happens if, in the future, everything breaks humanity’s way.
ARIA:
We’re speaking with visionaries in many fields, from art to geopolitics, and from healthcare to education.
REID:
These conversations showcase another kind of guest. Whether it’s Inflection’s Pi or OpenAI’s GPT, each episode we use AI to enhance and advance our discussion.
ARIA:
In each episode, we seek out the brightest version of the future and learn what it’ll take to get there.
REID:
This is Possible.
REID:
We’ve spoken about education before on the Possible podcast with previous guests, Ben Nelson and Ethan Mollick. While those conversations were insightful and far-reaching, they focused on higher and continuing education. That’s why today we’re discussing K-12 education.
ARIA:
There are many ways that AI can be applied in the classroom beyond students writing essays using AI chatbots. Since 2018, Singapore has implemented a national online learning platform that incorporates AI elements to adapt to students’ different learning styles. And South Korea is on its way to implementing AI coursework in its national curriculum across all grade levels by 2025.
REID:
There is potential to incorporate AI even further. It’s estimated that 40% of teachers’ time is spent doing tasks like grading or lesson planning. AI can help teachers with these vital but repetitive tasks and leave them with more time to develop relationships with students. AI can also help students in the form of chatbot tutors.
ARIA:
We’ve seen the internet widen access to education, and some are wary of AI entering the classroom, but there’s no denying that there’s a transformation coming in education brought about by artificial intelligence. Today’s guest is no stranger to the ways that technology can transform education.
REID:
Sal Khan is the founder and CEO of Khan Academy, a nonprofit educational organization that offers free lessons in math, science, and humanities, as well as tools for parents, teachers, and districts to track student progress. Khan Academy is currently piloting an AI guide called Khanmigo that is a tutor and teaching assistant.
ARIA:
So we sat down and we talked to Sal Khan about—I wanna say the future of education, but it’s really the present. I mean, Khanmigo is being piloted in, you know, classrooms and districts across the country. And he sees enormous changes coming in the next six months, the next year, the next five years.
REID:
And Sal is just literally the kind of—how do we build the future education that elevates all humanity? And approaches that in a, you know, a thoughtful, scale, humanity-driven way. Now we learned all kinds of things. A little bit about heavy metal. But of course, the real driver is how these AI chatbots—Khanmigo, others—can help elevate humanity. And you know, education is obviously one of the classic ones by which we, you know, elevate ourselves, elevate our children, elevate our communities.
ARIA:
And I think one of the things that people worry about with AI and education is taking the humanity out of it and like, “Oh, you’re just gonna be sitting in front of a screen and sort of passively watching.” And it was so interesting to learn about, in Sal’s school that he created, he was saying that even with remote schooling, they only watch a screen for, you know, an hour-and-a half or two hours a day. The beauty of AI is that it can actually unlock more personal time, more relationships. And Reid, I think, as you said, so many teachers go into education not because they’re dying to grade papers all day, but because they wanna have a meaningful impact on young people’s lives. And so giving them more opportunity to get to know their students, to be one-on-one with students, to have those interpersonal relationships—again, paradoxically, AI and technology can bring us more human interactions and relationships, which is so beautiful.
REID:
And here’s our conversation with Sal Khan.
REID:
Sal, it’s awesome to see you again. One of the pleasures that I get of doing these things with friends is I get to ask questions I haven’t asked before. I learned from Bill Gates’s podcast that you were a skilled drawer and you were a singer in a heavy metal band when you were young. Like, tell me a little bit more. What other secret talents do you have? Like is, you know, a heavy metal band in Khan Academy’s future? You know, what’s going on?
SAL KHAN:
I will take some credit for the drawing side. That is a talent that I probably haven’t nurtured as much as I should have. But yes, in high school I was the kid that could draw well. And I was our school artist, and, you know, I could draw portraits. It was a good thing. The heavy metal lead singer, I would not call a talent. It’s something [laugh], I would say. But yeah, I was, I had little aspects of my personality that were probably a little bit angrier back then, a little angstier. And I had an outlet, and as the lead, almost lead yeller or growler of a heavy metal band.
ARIA:
Very nice. I love it so much. Creative pursuits alongside all the rest. So I’m sure you’ve gotten asked this question a million times, but the origin story of Khan Academy is so lovely. And for those who are listening to the pod and they don’t know, you know, it was born out of you tutoring your, I think then-12-year-old cousin Nadia in math. And at the time you were at a hedge fund in Boston and she was in New Orleans, and so, you know, you were doing it via video, and this experience grew into this massive business. Like, what does Nadia or your family members, like, what do they think about what Khan Academy has grown into from that little seed?
SAL KHAN:
They’re proud of it. And, you know, their mom, my aunt, my uncle’s wife, you know, she’s always, “Sal, you know, thank you so much. It really helped the kids. It put them on a trajectory.” I’m like, “No, thank you [laugh]. If you didn’t let me tutor your kids, who knows what I’d be doing right now?” So yes. And they’re doing great. Nadia’s about to get married. She’s finishing up a PhD in clinical psychology. And her two younger brothers who are, you know, they both ended up going to MIT and now are software engineers. So they, yeah, they’re doing well.
ARIA:
That’s awesome. And so, you know, Khan Academy launched 15 years ago. There has been a lot of, you know, evolving in ed-tech since then. What do you think has changed now versus 15 years ago? And is that change accelerating, or what would you say?
SAL KHAN:
Yeah, let’s go back 15 years ago. You know, ed-tech was not an exciting place [Laugh] 15 years ago. I think most venture capitalists did not even view it as a thing to invest in. And look, one of the reasons why I set up Khan Academy as a not-for-profit is: I also didn’t think it was a place [laugh] where market forces were necessarily leading to good outcomes, or at least to outcomes that were aligned with our values. In a weird way, I think—and I don’t wanna be too self-aggrandizing of Khan Academy—but I do think in a weird way, back in 2010, 2011, we did show that, “Hey, there is a market need here, and there are ways to use technology to create scaled solutions that people want.” And we’ve heard, you know, some of the folks who started the MOOC shortly after, they say, “Oh yeah, we saw Khan Academy.”
SAL KHAN:
They said, “Let’s apply this to college.” And there were a lot of other for-profits that also said, “Oh, it looks like there is.” And so the dramatic difference is you actually see a lot more investment. And I think broadly speaking, that’s a good thing. If you have more resources trying to solve a problem, whether it’s for-profit or nonprofit, that’s a good thing. But if you fast forward today, yeah, ed-tech is kind of a pretty crowded space. There’s a lot of ed-tech players. And back in 2008, 2009, it was kind of crickets.
ARIA:
And would you go back and still start an NGO? You know what I mean: Do you think that was the right choice with hindsight?
SAL KHAN:
I think for what Khan Academy is still trying to do, 100%. You know, my thought process for it—and it was a little delusional back then—actually, it was very delusional back then. It’s still a little delusional—you know, as you mentioned, my day job, I was an analyst at a hedge fund. So it’s not like I don’t believe in capitalism. [laugh] I don’t believe in free markets. Arguably that’s all I was doing back in the day. But I also saw, working as a hedge fund, you know, I was the annoying guy that would call up public companies, get on their investor calls and like, you know, hold their feet to the fire of, like, what their earnings are going to be this coming quarter. And on one level, that was my day job.
SAL KHAN:
But on another level, I was really disappointed by how most people had to think on that short-term. And I almost became a student of capital structure. And I saw that, whether for-profit or nonprofit, if you have a very vision-focused founder or set of founders who really believe in what they’re doing, they can do very mission-focused things. But as soon as they, their control gets diluted or there’s a change in control, the organization is very likely to not stay true to that same mission. And when I thought about Khan Academy in those early days, I thought about all of the people who were sending me letters and saying how they were using it. And it was really making me feel good that, you know, somebody who’s deployed in Iraq is able to use this resource to feel confident to go to college. Or there’s an orphan girl in Mongolia who’s using it to, you know, learn algebra.
SAL KHAN:
And I thought about almost every permutation of a for-profit I could imagine would’ve probably had to close off some of those doors for some of those people. And I read a lot of science fiction, and I imagine: Well, you know, a homerun in the for-profit space, it’ll turn into the next Google or the next, you know, Microsoft or something, which is great. You can do a lot good there if you do it well. But I was like, well, what’s a homerun in the not-for-profit space? It’s like, well, what if Khan Academy can almost be like a Smithsonian or an Oxford that can be true to that mission for generations, and actually could do it on a completely, on a whole new scale? Even when it was, when I was just one guy, I used to show philanthropists that Khan Academy—this was back in 2008.
SAL KHAN:
We already had more usage, I believe, in a month or in a year at that, back at that time, than Harvard had in its whole history. And I was just like—and we’re still growing. You know, now we’re at least a hundred times, or maybe a thousand times, bigger than we were back then. So yeah, that was kind of the dream. And, you know, you fast forward, there’s several case studies, Harvard Business School case studies, on “Should Khan Academy be a for-profit, not-for-profit?” And the same issues always surface: Access to capital, access to talent. There’s some, I would say, stereotypes that not-for-profits can sometimes be a little bit less nimble or a little less innovative. But where we are now, access to talent, we can’t pay what a lot of the other folks in tech are paying, but we pay well, very well, for a nonprofit. You know, it’s sometimes, I think, a blessing where you have to deal with scarcity.
SAL KHAN:
It makes you a little bit more nimble. It makes you really focus on what you do. And, you know, and whenever someone, I think, even implicitly implies that not-for-profits can’t be as nimble or competitive, and I always remind our team: We’re not competitive for our own egos; we’re competitive for the mission. But whenever someone, I was like, “Well, we’re gonna prove you wrong.” And I would tell people, you know, “pound for pound, I think the Khan Academy team can out-innovate for sure almost anyone in edtech. And I think even, even beyond.” [Laugh]
REID:
Well, I think the results are really great. And I think that provides an excellent bridge to starting to talk about the future, which is Khanmigo, you know, an AI chatbot tutor you built, describes itself as whimsical, eccentric, brimming with a sense of wonder. Those are not necessarily the words that many people would use when, you know, describing an AI assistant. Reminds me, of course, deeply of how we think about Pi, the Inflection, you know, company that I co-founded with Mustafa and Karén. So what drove you to create Khanmigo? Why do you think it’s important to highlight the human attributes? Where do you think this going?
SAL KHAN:
Yeah, it’s fascinating. It’s pretty much all I think about these days. You know, if, and it actually, you know, we talked about the origin story of Khan Academy, and in the past I told that story just as like, “Hey, this got started as a family project, et cetera.” It’s interesting. Now, I think it’s very relevant to what we are actually trying to do with AI in a lot of ways. As we just talked about, it started with tutoring family. And back in 2005, 2006, the first thing I did—a lot of folks think about the videos—but the first thing I did is, I need to give my cousins more ways to practice, get immediate feedback, learn at their own time and pace. So I created the software that would create problems at the time. Now we just have a very large item bank.
SAL KHAN:
And then a friend suggested I make videos. I made videos so that my family could get on-demand help. And what I really was trying to do was scale up what a tutor would do, scale up personalization. And I was able to do it with somewhat, in hindsight, rough instruments—like just self-paced practice software or videos and things like that. Fast forward to last year, and OpenAI reaches out to us, and they say—you know, this is well before ChatGPT came out. And as everyone—you know, ChatGPT is built on GPT-3.5. OpenAI comes to us the summer of 2022 and says, “Hey, we’re working on our new model. For various reasons, we’d love to talk to y’all once it’s done training.” I, from a, just a tech-nerd point of view, I was intrigued because I was like, “Oh yeah, GPT-1, 2, 3, you know, they’re getting cooler every time.”
SAL KHAN:
But from an education point, I was super skeptical. Because even GPT-3, very cool, but it really didn’t seem to have a handle on knowledge. You really, you go two interactions in, and you’re like, “Okay, it’s not really understanding what’s going on.” But when they first showed us, and then gave us access to, GPT-4—even though it’s also not perfect—we started realizing, wow, a lot of these ideals of scaling, say a tutor for every student, or a teaching assistant for every teacher, this isn’t science fiction anymore. [Laugh]. Like, this is actually doable in our lifetime. And now. And we started having arguments—or debates, I’ll call them—within Khan Academy. And this is before ChatGPT came out, everything, we were under an NDA with OpenAI at the time, a non-disclosure agreement.
SAL KHAN:
And half of us, and I was probably more on that half, were like, “This changes everything. We have to go all in on this.” And then the other half of the team rightfully was saying, “but hey, look, you know, people trust us. This stuff can still hallucinate. The math is infamously bad. Do we really want to be the people who just run out front on this?” And what I told the team is, “Look, those are real risks. That’s legitimate. And we also have to worry about bias. And we’re. going into under 18, which increases the scrutiny even higher.” But I said, “Look, let’s write all of these risks down, and let’s turn them into features.” Like let’s not use it as a reason to shy away. And so as our team actually started to get really aligned that wow, we can actually turn all of these risks into features that in some ways can almost mitigate some of the risks that exist on the internet generally, even before you talk about generative AI.
SAL KHAN:
So we, you know, we have an AI that moderates the conversations, actively notifies parents or teachers. If we’re talking about an under 18 user, if there’s something that seems to be going in a not constructive discussion, all interactions for under 18 users are transparent. Khanmigo will not cheat. This was well before ChatGPT came out, but we’re like, okay, cheating is a clear use case here, but it will work with you socratically. And a lot of it is anchored on the Khan Academy content. So that’s been able to minimize a lot of the hallucinations. And then we do a lot of, we’ve done a lot of work on the math side to get a lot better. We were even part of the fine tuned training of the underlying GPT-4 model. I was personally, I probably spent about 20 hours giving it feedback [laugh] on how it could respond to math and tutoring questions better.
SAL KHAN:
But that’s been our take and, you know, when ChatGPT came out, and at the end of November in 2022, I remember I was like, “what’s going on here?” And I slacked the OpenAI leadership team, and I was like, “Hey, hey guys, you have us on our NDA, and you just launched something [laugh]—like, what is this?” And they told me, “oh no, no, you know, this isn’t GPT-4. We just put a chatbot on front, on top of GPT-3.5,” which had been out for six or seven months. And the whole world just kind of, you know, woke up to this. And I was really worried because this was not built for education. It was cheating, and it was cheating in a bad way. It would make up things. Its math was infamously bad. You start seeing school systems ban it. I’m like, “Oh no, we’re putting so much effort into this, and now people are gonna throw out the baby with the bath water.” You fast forward to March of 2023, it was almost a blessing. Because people I think at that time had realized that, okay, this isn’t a technology that we should ignore. If only someone put guardrails on it. If only someone thought about it from an education lens. And then we were able to say, “Yes, someone has. Here’s Khanmigo.”
ARIA:
I tried Khanmigo with my eight-year-old son, Luke, and he has Famous Friday at his school every week. So we use that and Khanmigo’s “chat with a historical figure” feature as a jumping-off point. Okay. Luke, we are looking at this list of famous people to chat with. Who do you wanna chat with?
LUKE:
Albert Einstein.
ARIA:
What are you gonna ask Albert Einstein, Luke?
LUKE:
Can I first just like introduce myself?
ARIA:
Yeah, introduce yourself. Great idea.
LUKE:
Okay. Hi. I am in elementary school. Um, I want to ask him: How does energy equal matter squared?
KHANMIGO AS ALBERT EINSTEIN:
Ah. You’re referring to my most famous equation: E = MC^2. This equation states that energy, E, equals mass, M, times the speed of light, C, squared. It’s a fundamental principle of my theory of relativity. The equation shows that mass and energy are interchangeable. They’re different forms of the same thing. If mass is somehow converted into energy, it also shows how much energy would result and vice versa. Now let me ask you a question: How do you think this concept might apply to the world around us? For example, consider the sun, our primary source of energy.
LUKE:
Um, because it can also turn into mass because, I think that the thing the sun is made out of—plasma—is a form of matter and is matter and, um, is matter. Yes, matter is mass.
ARIA:
So I type that into Khanmigo, and let’s see what they responded.
KHANMIGO AS ALBERT EINSTEIN:
Ah, you’re on the right track. Indeed, the sun is primarily composed of plasma, which is a state of matter. The energy produced by the sun comes from a process called nuclear fusion, where hydrogen atoms combine to form helium. This process converts some of the mass of the hydrogen atoms into energy, according to the principle of E = MC^2.
ARIA:
I think that what not everyone realizes, and I would encourage everyone to go watch your TED Talk, is like the basic idea that one-on-one tutoring can move, you know, a bad learner to a good learner, and a good learner to a great learner. And so if we could just apply that to everyone. And so one of the questions I get, to your point about, “oh, I wish,” if you’re a classroom teacher today, if you’re working, you know, K-12 in the public schools, what should you do? What should your interaction with AI be? Should it just be to use Khanmigo? Like, what would you tell all these teachers who have questions about, you know, how their profession is gonna change?
SAL KHAN:
I have a short-term answer for that question, and I have a longer-term answer. I think in the short run, I think it’s always valuable to use class time to do more active learning. We’ve been preaching this since the beginning of Khan Academy. It’s kind of beside the point when you spend class time to lecture, and then you make students do all of the active learning on their own time and pace, and they’re not getting any support, they’re not getting immediate feedback. We’ve always said this in math classes, and I would say this now in writing classes: Like, it’s actually nice to treat your class more like a writer’s workshop, have students write in class. And then you also have control about what tools they may or may not be using. So that’s one extreme. Now the other extreme is—and I think maybe this is more appropriate if we’re talking about older students, college-aged students, high school—the world is using these tools.
SAL KHAN:
And so I think it is healthy to have some projects where you tell the students, “go off, use whatever tools you need, but be more ambitious.” So maybe in the past in a business class in college, they might’ve said, “Hey, I want you to write a business plan.” Now the professors say, “I want you to start a business.” [Laugh]. And use and use whatever tools you need to write the business plan.” But it’s gonna be your plan. So it’s not like you can just check out and, you know, have, have an AI do whatever it wants with it. Where we’re going—and it’s not far, it’s about a year away: What we’re trying to do at Khan Academy, especially with Khanmigo, is, I would say, a middle path that is actually probably going to be the, I think, the dominant path.
SAL KHAN:
So we want students of all ages to still be able to do independent research, to be able to write on their own. But then there’s the question of: How can you do it in a world where these tools make it so easy, and it’s not really plagiarism or it’s not detectable plagiarism in the same way. And what we are doing is, imagine a world where a teacher—and a teacher can already use these tools to do things like create lesson plans and rubrics for assignments and assignments. So that’s already happening on Khanmigo, but a teacher can do that. And then through the AI make the assignment to the students. So when the student goes onto the platform, Khanmigo tells them, “Hey, we gotta write this paper on the following topic.” And the teacher can decide how much support that the AI can give, but it’s not gonna do the essay for the students. It’s going to work like an ethical, thoughtful writing coach alongside the student.
SAL KHAN:
And they might riff a little bit on what a good thesis statement it is. And we’ve already built this tool where it can give feedback, where it’s highlighting parts of the passage, and it’s starting really threads—the way if we were collaborating on a Google Doc, you would highlight: “Hey, Sal. I don’t know if this wording’s exactly right,” and we’d go back and forth. So it’s already doing that, giving feedback. They can iterate. It can also give feedback on, based on the teacher’s rubric that probably was done in collaboration with the AI: “Hey, this is where you probably are.” And then when the AI, when the student’s ready to submit, and the AI reports back to the teacher, they’re not just reporting back the output. You know, when we, when all of us did essays as kids, we just handed in the essay. And honestly, even back then, the teacher did not know who wrote that essay.
SAL KHAN:
It might’ve been our friend. Even before generative AI, if you do a search for “write my essay,” it’s not hard to find seemingly legitimate, but clearly not legitimate, operations. These essay mills that will write your essays for $5 a page. Things like ChatGPT just lowers the cost of that. But what happens now is that the AI, Khanmigo, when it reports back to the teacher, it’s not just gonna report the output. It’s gonna say, “Hey, myself and Aria, we worked on this for about four hours. She had a little bit of trouble initially with a thesis statement, but we got there. I had to, you know, some of the sources she was using didn’t seem credible. So we give a little bit of feedback. By the way, teacher, here is the whole transcript. And her writing is consistent with what we’ve seen her produce in class when we work together. So I feel pretty good.”
SAL KHAN:
Now, on the other hand, Reid goes out there, uses ChatGPT, or uses one of these essay mills and copies and pastes that into Khanmigo. Khanmigo is going to tell the teacher, “I didn’t work on this. This is shady. It just showed up.” And so not only will AI support students better and give them better or faster-cycle feedback, which you need for anything to be better, but I actually think it undermines decades, if not centuries, [laugh] of the potential for students to cheat in that way.
REID:
One thing that you already covered a little bit of, but I think it’s also important: It’s very natural to talk about how this amplifies students in all kinds of ways. Talk a little bit about also how it amplifies teachers, right? Because both sides are being amplified in creating this revolution in education. And it’s partially that, you know, the teacher’s like, “well, I don’t want to change the way that we’ve been doing it for the last 50 to 80 years.” And you’re like, “No, no, no. A) You’re gonna have to, but B) It’s going to be magical.”
SAL KHAN:
The more I think about it, this is the dimension that I might even—you know, it’s hard to pick a favorite child, but I’m, I might be more excited about—because the pitch to teachers, Khan Academy’s been making this pitch, and pretty much every edtech player has been making the pitch for the last 10 years. They’d say, “look, we’ve created a great tool. We have a bunch of efficacy studies, and if you adopt it in this way, your students are going to accelerate by 20, 30, 40%.” And teachers, I’ve seen this with the last 10, 15 years at Khan Academy, they’re like, “Oh yeah, we really believe in personalization. We believe in differentiation. We wanna address their learning gaps.” But they’re already overworked, and you’re giving them one more tool to learn, one more thing to implement. And so it’s hard, and it’s, it spreads them even thinner.
SAL KHAN:
What’s powerful about generative AI, we still are gonna be able to make those cases. And it’s early days but we’re already running the efficacy studies, et cetera. But one dimension of it, which is acting as a teaching assistant, directly does not add work to teachers, but actually can take work away from them. The average teacher spends 10, 15 hours a week doing things like writing lesson plans, grading papers, writing progress reports, developing rubrics. Generative AI can do all of that. Now, I’m not gonna advocate that you should somehow take the teacher out of the loop, because it’s not perfect. The teacher should be there. But generative AI can make it a lot easier for them to tweak the lesson plans. To say, “Hey.” They construct the rubric together, and based on it, “Here’s a first pass on what I, Khanmigo, think the student scored…” But if we can take that 15 hours a week and make it an hour-and-a-half a week, that’s a massive gift to teachers.
SAL KHAN:
And that’s, they need that energy for themselves. Hopefully we can stop the attrition that we’re seeing in the teaching profession. And then that’s more energy that they have with their students. And, of course, every teacher, the great majority of teachers, want to differentiate more, serve their students better. We’re already seeing that in classrooms, where there’s about 50,000 teachers and students in real classrooms. You know, we have a lab school, but we’re also talking about places like Newark, New Jersey. We’re talking about Hobart, Indiana. And the teachers are telling us 1) they love the teaching assistant aspect of it. But they’re like, my kids are, the younger kids, are learning to ask questions better that they were afraid to ask. They’re starting to realize where their kids were having questions. Because most kids aren’t bringing them up. And they’re able to save time for themselves and feel like they’re supporting their students better. So it does feel, Reid, to your point, like it is magical, and even though this is a more advanced technology than anything in the last 10, 15 years, in some ways it’s a more natural one, which I think will make it cooler.
REID:
The other thing, one is the teacher; people don’t realize that it’d be great for the teachers. Teachers can spend more time with the students. They can spend more time–less on rote work, you know helping, you know, part of why the great teachers get into this is because they love to help children, you know, and students. And there’s like more time doing that. The other one is also, which we hinted at a little bit, which is, you know, wealthy families can always, of course, afford tutors, afford other kinds of things. I know it’s very early days, but have you seen any of the kind of Khanmigo expanding that aperture more to the various, you know, kind of more impoverished communities and so forth that can go, “Oh, now I get a tutor too,” and that kind of elevation of talent, that inclusion, in great outcomes? And if there’s anything in particular, please share.
SAL KHAN:
Simple answer is yes. And it’s still very, very early days. The North Ward of Newark, New Jersey, that actually started intensely using Khan Academy pre-Khanmigo, November of last year. And it’s actually turning into, you’re gonna hear more about the North Ward of Newark, New Jersey. They have some amazing leadership at the ward level, and it’s now expanding to the district level. So when Khanmigo came out, they were one of the first places where we said, “Hey, do y’all want to try this out?” And they did. And we’re seeing exactly that. You know, we have video footage that I’m now showing at some talks I give, where it’s kids in Newark, who are not rich kids by any stretch of the imagination, who are saying, “Yeah, it feels like I have a tutor sitting next to me. It feels like I have a friend, you know, and I feel really supported.”
SAL KHAN:
And Khanmigo still is not going to be able to do everything that a, you know, the upper-middle class or the affluent can hire someone to do. But it is there when you need. In fact, in that way it’s better than even a $50-an-hour or a $100-an-hour tutor because it’s there in the middle of the night. It’s there right, when you have the problem, you don’t have to be embarrassed about it. And I think in the next year or two—and we are working on this, I’m sure others are too—it’s going to transcend, in fact, we are gonna make Khanmigo likely transcend Khan Academy. So it’s not just when you’re doing your Khan Academy work; it can be your guide when you’re doing anything. But also, you know, proactively make sure you’re engaged.
SAL KHAN:
So right now it’s helping you in your stock. It’s helping you explain the concept in a Socratic way. But when I realized what I was doing with my cousins almost 20 years ago, a lot of it was me calling up their mom and saying, “Where are they? You need to get rid of that extracurricular because they’re falling behind in math. And that’s what they have to focus on right now.” So I actually think the future—and when I’m talking about the future, I’m talking about like in the next year or two—it’s going to be, it’s going to give a holistic picture. We already have—Khanmigo can report back to the adults, to teachers, saying, “Hey, this is what I’ve been working on with your students. And by the way, did you know that Aria’s really into soccer?” And when we talk about—we’re actually giving a sense of memory already to the AI. And it’s taking notes, which are transparent to the students and the teachers. They can modify it. So it’s, you know, it’s not some opaque black box. But I think, yeah, it’s amazing how fast the technology is advancing. And it’s really just an engineering question of putting in the pieces in the right way that feel, to your point, Reid, magical. But I think that’s gonna happen in the next three to five years.
ARIA:
I love the feedback point. I was at a parent-teacher conference for my eight-year-old, and his third grade teacher said, you know, “Don’t help your kids with homework because I need to know where they are. If I know where they are, I can customize lessons for them.” And so the more feedback a teacher can get, they don’t just have to rely on these, you know, analogue ways. And there’s so much research that says if, you know, a 10-year-old just has one caring adult in a school who knows about them, who asks them about soccer, who like that can transcend anything—so again, the learning is so critical, but also what it opens up teachers to be. You know, these real-life companions that get kids and know them and sort of ask them the right questions. So one question I had is that, you know, as many parents during COVID found out, remote learning was magical and terrible. And, you know, screen time; a lot of studies have shown that brain activity drops when you’re on Zoom calls, as opposed to in-person. Like, how do you think about that, and how do you keep people engaged in Khan Academy, Khanmigo, et cetera, when it is sort of intermediated through a screen?
SAL KHAN:
Yeah, and you know, I’ve said this before the pandemic, and I believe it even stronger during or after the pandemic: Screens, technology, they are, they can be good, they can be bad. And anything in excess is, even good things in excess, are not necessarily good. I think the pandemic was almost the worst-case scenario, where in a matter of weeks, schools had to take oftentimes in-person interactions that I would not advocate for—you know, I, we have a lab school, and it’s, we say it’s, “Hey, it’s not about lecture.” You can have teacher-directed dialogue or simulations, but ideally kids are working with each other. And it’s—when humans are together, they should be interacting. But we know that the status quo in most classrooms are still: kids sit passively, quietly. They’re not allowed to talk. You know, even if your friend says, “Hey, do you know how to do problem number three?”
SAL KHAN:
And you say, “Oh yeah, it’s really easy. You just have to remember…” You know, someone’s gonna, “shh, you can’t talk.” I’m thinking about it. They’re shutting down an interaction that’s incredibly powerful. Like, one student’s explaining their understanding, another student’s getting it maybe in a way that connects with them better than what’s going on in the textbook or the lecture. So you take that type of, I would say in some cases dehumanizing interaction, and then you put it on Zoom [laugh]. And you have someone stare at a screen that’s, you know, two feet away from their eyes for like four hours on end. No one would expect that to be a good thing. And so I think it’s all about thoughtful use of it. At our lab school, I tell people, “Come visit.” Even though the school will use technology more than the average school, you are also going to see, I’d be confident to say, at least five times the person-to-person interaction—physical person-to-person interaction, that you would see almost anywhere.
SAL KHAN:
Because the students are able to use these tools. And even actually, while they’re using tools like Khan Academy, we encourage them to support each other in ethical ways. And it’s freeing up the faculty to do more games, more simulations, more Socratic dialogue. You know, people are blown away by how collaborative our students are. Because when you’re doing mastery-based learning, it’s not me-against-you in a curve, and it’s not me-against-the-teacher either. Is that person gonna fail me or not? It’s all about all of us. It’s us against, or trying to get to, a threshold of learning. And we’re all on the same team trying to get to a threshold of learning. So it makes it more collaborative. And we did start an online high school with Arizona State University called Khan World School and almost wanna do that on purpose to show that there is a way to do this.
SAL KHAN:
Now ideally, those students, even at Khan World School, have in-person context; they could be in a pod, et cetera. But you could imagine if you’re in rural Alaska or someplace where you don’t have access to a really rigorous school, this gives you that. But what our implementation of it is, it’s probably about an hour, hour-and-a-half of screen time a day for high school students. And it’s Socratic dialogue. They’re getting into debates about issues of the day. So they look forward to it. They get into it. And then all the other interactions. They might, you know, they might be on message boards and Slacks with their friends. They’re getting organic interactions where they’re forming clubs. And then they do check-ins, kind of Oxford-style, with their faculty, where once a week, say, “Okay, what were your goals? Were you achieving?” And if you talk to these students, they feel very engaged, very seen, and very human. So there’s ways to do it.
ARIA:
Can you tell us a little bit about how you think of, sort of, the international stage of education? There’s also the question of global competitiveness, just from an education perspective. And we see, you know, the Korean ministry’s Future of Education Center, they’re embracing AI and including it in the curricula. Do you think the US is behind the rest of the world in this? Are we slow to adopt? Or are we, are we leading the pack?
SAL KHAN:
I don’t think we’re behind yet. And if, you know, if we have anything to do with it, [laugh] we wanna see the US really be—you know, I’m hoping if we’re talking a year from now, we’re talking about a million or two million students who are like formally using Khanmigo as part of their day-to-day work in their classrooms. Like, not even just at home and things like that. And if we’re talking in five or ten years, I hope we’re talking in the tens of millions. And now we’re talking about a good chunk of the US. So I don’t think the US is behind yet. And one thing, I always like to take a step back, because it’s very easy for us in the US. We look at PISA rankings, and we’re like, “Oh, we’re right in the middle of the pack now.”
SAL KHAN:
And there’s all these countries—first of all, some of that data isn’t real data. They’re not comparing every American student to every Chinese student. They’re comparing every American student to the kids in China that the Chinese government want to be taking the PISA exam. There’s a lot of kids in farms in Western China who have not seen a school in years and would do very badly on the PISA exam. So there’s that. I think, and the US is also a very diverse place. Like Massachusetts, for example, would be pretty much at the top of, if it was a country in the PISA rankings, it would be right up there with Finland and Singapore, et cetera. And it’s about the same size as Finland and Singapore, right? It’s actually a fair comparison. But the really interesting thing is we have delegations coming to Silicon Valley from Singapore, from China, from Finland, literally.
SAL KHAN:
And they’ll always stop by our offices while they’re touring Silicon Valley. They want a tour of our lab school, and they will tell me, “you know, what are we doing wrong? We are, we are kicking your butt [laugh] when it comes to test scores. But it feels like all the innovation is happening here.” And my point of view, and I, you know, this is my American exceptionalist side of me [laugh], I’m like, “yeah, we in America should not be trying to turn our school systems more like the Finnish. Or, I mean, there’s some good ideas. I don’t wanna say we should ignore those. There’s some good ideas we can learn from other folks. But all school systems are essentially adopted. You know, it’s famously the Prussian education system. It came outta the Industrial Age. It’s a very German system, an Industrial-Age German system of, you know, standards; move everyone forward at a set pace, apply some processes, start filtering people out along different tracks at different points.
SAL KHAN:
And what I’ve argued is we should have a more American school system. What does that mean? Well, America—and I was born here. I come from an immigrant family; my mom, she gets embarrassed, not embarrassed, I mean, she, you know, when I tell people that, you know, we grew up below the poverty line. I grew up in a single-mother household. And in America, that’s a mark of honor—that like, look, look where I started. And my mom says, “Sal, don’t say that. Don’t—we grew up, we have a very good family. Your grandfather was this, this and that. People are gonna think.” But to me that was always a sign, like, wow, in America, that’s a mark of honor. While in other parts of the world, you almost lean more on like, yeah, what my great-great grandfather did, or this or that.
SAL KHAN:
And in America, Silicon Valley, which I would argue is the most American part of America, if you take a risk, I mean, people—I mean, Reid’s one of those people, he meets them all the time. Like, you take a risk and you fail; as long as you failed well [laugh], you know, you didn’t fail. It’s a badge of honor. You, and then you can, and people are gonna, as long as you didn’t fail in a horrible way, in a way that’s embarrassing or in a way that showed some kind of incompetence, people are more likely to take a bet on you, not less likely to take a bet on you. And so there’s all this literature about growth mindset where, you know, people with a growth mindset do better. Failure should not be viewed as a negative.
SAL KHAN:
It should be viewed as an opportunity to learn. But the reality is, most education systems, especially outside of the US, have the opposite of a growth mindset. Like anything less than a failure is like shame for your family. It’s not like you should go hide. And so I, you know, this is why, like, things like mastery learning—at Khan Lab School, if a student doesn’t master it the first time, we say, “Hey, you gotta keep trying that.” That’s actually what growth mindset is about. And that’s what entrepreneurial thinking is about. “Hey, you failed the first. Good, good try. Like, keep working on it.” So that’s what I tell the Finns. That’s what I tell the Singaporeans. Like, you have to create mechanisms so that you get multiple shots on goal and ideally use more of your class time for innovation, for entrepreneurship, for challenging the norm, not—
SAL KHAN:
You know, in some cultures, and I think the US actually could learn a little bit of this, I don’t, I think we’ve gone too far not revering educators. I think in some other cultures, the educator can do no wrong and the student is afraid to question the educator. I think the ideal is revere the educator, respect the educator, but both parties should be willing to question things. Be willing to question assumptions. And that’s what’s going to lead to innovation. And so that’s, you know, so the US definitely still has to improve our test scores, et cetera, et cetera. And that’s, but we should never lose what really makes, I think, this country special, and bring that more into the education system and then spread it globally.
REID:
Yep. Disruptive innovation, ask questions, and it’s a team sport we do together to improve in the world. We’re gonna see rapid progress from Khanmigo. What are some of the things that people should be thinking, not just of, “Oh, I can go play with it right now and see.” But I’ll be thinking about like, what are these AI personal intelligences, co-pilots mean for how we’re all going to get, kind of, better companions, team sport, as we’re going into our educational journeys?
SAL KHAN:
Yeah. And every time I think an hour about it, I’m like, “Oh, we could do that too.” And I’m sure there’s a lot of folks thinking in this way. There’s some obvious things. You’re gonna be able to converse with AIs like Khanmigo. I think in the next two or three years, you’re gonna be able to have video conferences with them. It’s going to feel very real. You know, as I mentioned, we’re already working on memory. Memory is fascinating. You know, a lot of people have disparaged the gen AI saying, “Oh, it’s just a, you know, a thing that’s just trying to predict the next word.” I’m like, “Well, what are you doing?” You know, I’m not—you know, I meditate a lot. And a lot of the art of meditation is learning to disassociate yourself from your mind and to observe your thoughts and realize that your thoughts are not you.
SAL KHAN:
And when you start to do that, you realize that your thoughts are very much like a large language model. It’s kind of a series of competing large language models that are all feeding off of each other. And then you are this awareness [laugh] that is like along for the ride on some level. So I think you’re going to see this become—I’ll give more, you know, we talked about the writing example; it’s gonna be used more for feedback, more for teaching assistants. I think you’re going to see—oh, and the reason why I started talking about this whole thought thing, the memory aspect of the AIs, which right now is, you know, we’re doing the application layer; who knows if future models, they somehow put it into the base model somehow?
SAL KHAN:
But the, you know, this notion of memory where you’re starting to realize that, you know, we’re building things that feel a lot like what we think dreams are. Or what, you know, why we sleep—you’re trying to consolidate your memories from the day because you can’t remember everything. And you start to realize that, oh, you have to consolidate memories one way if you wanna report back to a teacher or a parent. You have to consolidate memories in a different way if you’re trying to learn about how to motivate a student. You have to consolidate memories in another way if you wanna be able to refer back to a conversation that you had two weeks ago as an intelligent AI. So that, I find, from an engineering point of view, it’s gonna make it super cool and magical.
SAL KHAN:
But even from a “who are we?” point of view, [laugh] it’s really interesting. I think you’re gonna see, you know, these AIs done well—I know this can sound a little creepy—but done well, they’re gonna be able to have a long lasting narrative. In fact, there was just a study that some of the top-performing school districts are the ones where the teachers have multiyear relationships with the students, where it’s not starting from scratch every year. Ideally that’s happening in the physical realm. But if you have an AI that is, it’s the same AI that’s working with you in multiple contexts from year to year across multiple classes, then you have that continuity. And then when you have a new teacher, the teacher can ask the AI like, “What should I know about this kid? Like, you’ve been working with them for five years.
SAL KHAN:
Like, tell me more. Are they also doing, you know, are they also struggling in math? Because right now they’re writing is a little bit—,” and the AI can talk about that. Like, what’s their passion? So I’m super excited about that aspect. And, you know, I think you’re gonna see, as I mentioned, Khanmigo transcend Khan Academy. You can imagine an eight-year-old surfing the internet. Very scary proposition right now. I have a nine-year-old, a 12-year-old, and a 14-year-old. Right now, parents in schools, they have these very blunt instrument tools like GoGuardian. And no insult to them. I mean, they’re using, but they’re using 20-year-old technology to essentially filter out certain sites are okay, certain sites aren’t. I think there’s gonna be a future where the AI is going to be alongside you. It can help you get context. For younger students, it can actually remove context.
SAL KHAN:
You know, even a legitimate site, if you’re on New York Times, but it’s about war or sex crimes, you don’t want your nine-year-old reading that. It can change it. It can change the—I mean, you know this stuff can code. In some ways it can be a guardian angel on, you know, the variation that is our internet. That we, that no one really knew how to, how to—or even for any of us, like, “Hey, you know, you’ve been, you’ve been stalking your ex-boyfriend or ex-girlfriend on social media for the last half hour. How do you feel about yourself right now? Maybe we wanna get back to what, you know, what you were here to do.” I think we could all benefit. It could also be a therapist. I think we could all benefit from something that has our best interests. As we know, there’s already state-of-the-art AIs that are trying to feed us information and ads, et cetera, that have a very different objective function than our mental health and our happiness. Imagine if we have AIs on our own side that are, that are, that are fighting for our mental health and happiness.
REID:
We’re now gonna go to our rapid-fire section. Is there a movie, song, or book that fills you with optimism for the future?
SAL KHAN:
You know, I just rewatched “Contact” with my kids last night and I was like, “That’s a great movie.” There’s other books. I mean, Foundation, I’ve cited many times as one of the inspirations for Khan Academy. You know, and I’ve said, “Instead of preventing a Dark Age, what if we can help bring about something that will make today look like a Dark Age?” because we’ll be in such better place.
ARIA:
I love it. So is there a question that you wish people would ask you more often?
SAL KHAN:
Yeah, we touched on it a little bit. But I think a little, a little bit about like, what’s the point of all of it? [Laugh] Like, we can get very caught up in like the numbers or, you know, what leads to people’s income or this or that or success in the marketplace. But like, what’s the point of it? What’s all the meaning behind all of it?
ARIA:
I think we’re gonna have to have a whole second episode.
REID:
Yeah, exactly. I would love to ask you those questions in depth. So we’ll have to come back to that. Where do you see progress or momentum outside of your industry that inspires you?
SAL KHAN:
I’m, you know, and look, this might be a little cliché: I’m reading Walter Isaacson’s book about Elon Musk and I know, you know, y’all and we, both know him. And I know there might be dimensions of him that not everyone agrees with. But the—I find it very inspiring to see, almost through act of will, you know, industries that I thought no one was going to be able to touch—for, you know, once again, in our lifetimes—are already, you know, whether it’s electrification of cars or space exploration or interfacing between, you know, the mind and the digital world, yeah, it inspires, when I see things like that. I know he’s not the only person doing it. Other people are as well. But yeah, it inspires me to say like, “Okay, let’s not play small ball here. Let’s, let’s think big.”
ARIA:
Well, on that note, could you leave us with a final thought on what you think is possible to achieve if everything breaks our way in the next 15 years? And what’s the, what’s the first step in that direction?
SAL KHAN:
If everything breaks our way, you know, I’ll start, I’ll start “small” on the education side. I think in 15 years, you literally will have every child—not child, every learner on the planet—has access to some form of what we could call “world-class education.” And I’ve been saying, you know, Khan Academy’s mission is free world-class education for anyone anywhere. If you asked me even three, four years ago, and you said, “How are you gonna give every child rich, Socratic dialogue? How are you gonna give every child the ability to role play in a simulation, or to explore ancient Rome?” I was like, “Yeah, I’m not sure exactly how to do that right now, but Khan Academy’s going to be around well beyond me, so one day, you’re going to have AI, and you’re going to be in an immersive world, and you’re gonna have a link straight into your brain. You’re gonna have these lucid dreams where you’re in Rome. Whatever.” That’s likely to happen in the next 15 years, and it’s likely to be very accessible.
SAL KHAN:
So I’m excited about that. I’m excited about AI actually helping potentially to facilitate conversations between people and be that trusted moderator in certain cases. I think a lot of the rough edges of the internet, AI could actually, as opposed to—obviously it can amplify some of them with fake news and misinformation, all that—but I think that used well, we can mitigate those risks, and it can actually help us go back to the center on a lot of things. Be a source of thoughtfulness. One of my biggest fears of AI is what, what do totalitarian governments do? You know, right now they can already snoop every phone line, et cetera. But now they’re going to be able to use this to, to actually, you know, know what’s being said on every one of those phone lines, which is scary. And they could put sensors on the street, et cetera.
SAL KHAN:
But I think if everything breaks our way, there’ll, AI will actually allow us to even undermine that. And, you know, the misinformation campaigns can go both ways. You can have, you know, some of—you can create enough distrust in that, you know, dictator’s close…I don’t know if that’s the line of attack you want to go, but I hope that there’s ways that it can actually help undermine, kind of, this authoritarian type of control. Not 100% clear, but, and I’ll say one last thing: I think if everything goes our way, you know, there’s—we know it’s going to make us dramatically more productive as a society. And I think the big question is: Where does that productivity accrue to? And there’s one reality where all of us in Silicon Valley are the beneficiaries of it, but a lot of people are left out. And the only stable equilibrium there is some type of massive redistribution.
SAL KHAN:
And that’s not a—that, I consider that dystopian because people don’t want a handout. People want to have a sense of meaning. They want to be able to participate. So if everything breaks our way, the education element of AI is hopefully going to invert the labor pyramid and have most people be able to participate at, today, the top: Be a knowledge worker. Be able to be an entrepreneur. And for anyone who’s a naysayer, and sometimes I even naysay myself. I’m like, “Sal, is that realistic?” I remind—like, if you go back 500 years, only about 20% of people could read, and algebra was considered like this esoteric thing that [laugh] only a few people on the planet understood. Now we expect most people, you know, in places where we have education, most people should know these things. So today, what’s considered esoteric? How many people understand quantum physics? Or could start the next tech company? Or understand how a neural net works? I think there’s a utopian world where in 20 years everyone does. It’s just common knowledge. And then things are gonna be pretty good.
ARIA:
I love it. Sal, thank you so much for being here today. It was a delight.
REID:
And awesome. And Sal, as always, could easily talk for another few hours.
SAL KHAN:
Likewise. Thanks for having me.
REID:
Possible is produced by Wonder Media Network, hosted by Aria Finger and me, Reid Hoffman. Our showrunner is Shaun Young. Possible is produced by Edie Allard, Sara Schleede, and Paloma Moreno Jiménez. Jenny Kaplan is our executive producer and editor.
ARIA:
Special thanks to Katie Sanders, Surya Yalamanchili, Saida Sapieva, Ian Alas, Greg Beato, Ben Relles, Stacey Olson, and Little Monster Media Company.