Sal Khan on the future of K-12 education

What if every learner on the planet had access to a world-class education? This week’s guest has an impressive track record of using technology to make that dream a reality. Sal Khan is the founder and CEO of Khan Academy, an education nonprofit whose pioneering digital tools support K-12 students everywhere, along with parents, teachers, and school districts. In this episode, Sal discusses how technology widens access to education and creates more meaningful learning experiences. He also shares his experience helping train Chat-GPT and speaks to why, 15 years after launching Khan Academy, he has decided to go all in on AI. Plus, a special guest joins the podcast to demo Khan Academy’s new AI chatbot tutor, Khanmigo. Read the transcript of this episode here. For more info on the podcast and transcripts of all of the episodes, visit www.possible.fm/podcast. Topics: 5:26 - Hellos and intros 7:51 - What has changed in education technology in the past 15 years 13:24 - Khan Academy’s new AI guide, Khanmigo 18:59 - Khanmigo demonstration 21:37 - How teachers can incorporate AI into their classrooms 29:15 - AI in the classroom in North Ward, Newark, NJ 33:13 - How to make remote learning engaging 37:01 - How education technology in the U. S. compares to other countries 42:43 - How the use of AI in the classroom will expand in the future 47:28 - Rapidfire questions The award-winning Possible podcast is back with a new season that sketches out the brightest version of the future—and what it will take to get there. Most of all, it asks: what if, in the future, everything breaks humanity's way? Tune in for grounded and speculative takes on how technology—and, in particular, AI—is inspiring change and transforming the future. This season, hosts Reid Hoffman and Aria Finger are speaking with a new set of ambitious builders and deep thinkers about everything from art to geopolitics and from healthcare to education. These conversations also showcase another kind of guest: AI. Whether it's Inflection’s Pi, OpenAI’s ChatGPT or other AI tools, each episode will use AI to enhance and advance our discussion. Possible is produced by Wonder Media Network and hosted by Reid Hoffman and Aria Finger. 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. Special thanks to Katie Sanders, Surya Yalamanchili, Saida Sapieva, Ian Alas, Greg Beato, Ben Relles, Stacey Olson, and Little Monster Media Company.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:01Speaker 00

  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 said, well, how are you going to give every child rich Socratic dialogue? How are you going to give every child the ability to 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 is 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 going to have a link straight into your brain. You're going to 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.

• 0:51Reid Hoffman?

  Hi, I'm Reid Hoffman.

• 0:52Aria Finger?

  And I'm Aria Finger.

• 0:54Reid Hoffman?

  We want to know what happens if, in the future, everything breaks humanity's way.

• 1:00Aria Finger?

  We're speaking with visionaries in many fields, from art to geopolitics and from healthcare to education.

• 1:08Reid Hoffman?

  These conversations showcase another kind of guest, whether it's Inflection's Pi or OpenAI's ChatGPT. Each episode, we use AI to enhance and advance our discussion.

• 1:20Aria Finger?

  In each episode, we seek out the brightest version of the future and learn what it'll take to get there.

• 1:27Reid Hoffman?

  This is possible.

• 1:35Reid Hoffman?

  We've spoken about education before on the Possible Podcast with previous guests, Ben Nelson, Ethan Molick. 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.

• 1:52Aria Finger?

  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.

• 2:19Reid Hoffman?

  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.

• 2:42Aria Finger?

  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.

• 3:01Reid Hoffman?

  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 Khan Migo that is a tutor and teaching assistant.

• 3:24Aria Finger?

  So we sat down and we talked to Sal Khan about, and I want to say the future of education, but it's really the present. I mean Khan Migo is being piloted in classrooms and districts across the country, and he sees enormous changes coming in the next six months, the next year, the next five years.

• 3:43Reid Hoffman?

  And Sal is just literally the kind of, how do we build the future education that elevates all humanity? And approaches that in 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, Khan Migo, others, can help elevate humanity. And education's obviously one of the classic ones by which we elevate ourselves, elevate our children, elevate our communities.

• 4:21Aria Finger?

  And I think one of the things that people worry about with AI in education is taking the humanity out of it. And like, oh, you're just going to 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, hour and a half or two hours a day. The beauty of AI is that that 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 want to 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

• 5:04Aria Finger?

  with students, to have those interpersonal relationships. Again, paradoxically, AI and technology can bring us more human interactions and relationships, which is so beautiful.

• 5:17Reid Hoffman?

  And here's our conversation with Sal Khan.

• 5:26Reid Hoffman?

  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' 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?

• 5:51Speaker 00

  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, 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.

• 6:35Aria Finger?

  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, you know, 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?

• 7:10Speaker 00

  They're proud of it. And, you know, their mom, not with Auntie Maya, my uncle's wife, she's always, Sal, you know, thank you so much. It really helped the kids and put them on a trajectory. I'm like, no, thank you. If you didn't let me tutor your kids, who knows what I'd be doing right now. So, so, so, yes. And they're doing great. Nadia is about to get married. She's finishing up a PhD in clinical psychology. And her two younger brothers who are, you know, they, they both ended up going to MIT and now are software engineers. So they, yeah, they're, they're, they're doing well.

• 7:49Aria Finger?

  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?

• 8:03Speaker 00

  Yeah, let's go back 15 years ago. You know, ed tech was not an exciting place. 15 years ago, most, 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 that, where market forces were necessarily leading to good outcomes, or at least outcomes that were, were aligned with our values. In a weird way, I think, and I don't want to 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

• 8:45Speaker 00

  need here and there are ways to use technology to create scaled solutions that people want. And, and so, and, and we've heard, you know, some of the folks who started the MOOC shortly after they say, oh yeah, we saw Khan Academy, you said, let's apply this to college. And there was 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, 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, is, is kind of a, a pretty crowded space. There's a lot of ed tech players and back in 2008, 2009, it was kind of crickets.

• 9:27Aria Finger?

  And would you, would you go back and still start an NGO? You know what I mean? Do you think that was the right choice with hindsight?

• 9:32Speaker 00

  I, I think for what Khan Academy is still trying to do, a hundred percent. You know, my, 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. I, you know, my, 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. I don't believe in free markets. That's 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, ho, ho, ho, hold, 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. But on

• 10:15Speaker 00

  another level, I was really disappointed by how most people had to think on that short term. And, and I almost became a student of capital structure. And, 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.

• 10:43Speaker 00

  The organization is very likely to not stay true to that, to that same mission. And when I thought about Khan Academy in those early days, I thought about all of the people who are sending me letters and saying how they were using it. And it was really making me feel good that somebody who's deployed in Iraq is able to use this resource to get, feel confident to go to college, or there's an orphan girl in Mongolia who's using it to, you know, learn algebra. And, and I, and I thought about almost every permutation of a, of a for-profit I could imagine would have probably had to close off some of those doors for, for some of those people. And I read a lot of science fiction and I imagined, well, you know, a home run in the for-profit space.

• 11:28Speaker 00

  It's, 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, if you do it well. But I was like, well, what's a whole, whole, home run in the not-for-profit space? I was like, well, what if Khan Academy can almost be like a Smithsonian or an Oxford? That, that can be true to that mission for generations. And I actually could do it on a completely, on a whole new scale, even when it was, when I was this one guy. I used to show philanthropists that Khan Academy, this was back in 2008. We already had more usage, I believe in a, 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, and we're, we're still growing.

• 12:09Speaker 00

  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, that was kind of the, the dream. And you know, you fast forward, there's, 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, you have to deal with scarcity.

• 12:53Speaker 00

  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 going to prove you wrong. And I, and I would tell people, you know, pound for pound, I think the Khan Academy team can out innovate for sure, almost anyone in ed tech, and I think even, even beyond.

• 13:24Reid Hoffman?

  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 Khan Migo, you know, an AI chatbot tutor you built, describes itself as whimsical, eccentric, priming 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-battled with Mustafa and Karen. So what drove you to create Khan Migo? Why do you think it's important to highlight the human attributes? Where do you think this is going?

• 14:02Speaker 00

  Yeah, it's fascinating. It's pretty much all I think about these days, you know, Viv. 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, 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, 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, the software that would create problems at the time. Now we just have a very large item bank. And then a friend suggested I make videos. I made videos so that my family could get on-demand help.

• 14:47Speaker 00

  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, like just self-paced practice software or videos and things like that. Fast forward to last year and

• 15:06Speaker 00

  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, 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. 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, you go two interactions in and you're like, okay, there's something, some, it's, it's not really understanding what's, what's going on.

• 15:47Speaker 00

  But when, 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. Like this is actually doable in our lifetime. And now, and, and we started having arguments, 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, non-disclosure agreement. 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.

• 16:32Speaker 00

  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 worry about bias and we're, 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. 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. So we, you know, we have an AI that moderates the conversations, actively notifies parents or teachers if

• 17:16Speaker 00

  we're talking about an under 18 user, if there's something that seems to be going in a, in a not constructive discussion. All interactions for under 18 users are transparent. Conmigo 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 tune training of the, the underlying GPT-4 model. I was personally, I probably spent about 20 hours giving it feedback on, on, on how it, how it could respond to math and tutoring questions better. But, but that's been our take. And, you know, when ChatGPT came

• 17:58Speaker 00

  out in, 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 like, what is this? And, and they told me, oh, no, no, you know, this isn't GPT-4. We just put a chatbot on front, on, 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 going to throw out the baby with the bathwater. You fast forward to

• 18:43Speaker 00

  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 Conmigo.

• 18:59Aria Finger?

  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 want to chat with?

• 19:16Speaker 03

  Um, Albert Einstein.

• 19:19Aria Finger?

  What are you gonna ask Albert Einstein, Luke?

• 19:21Speaker 03

  Can I first just like introduce myself?

• 19:23Aria Finger?

  Yeah, introduce yourself. Great idea.

• 19:25Speaker 03

  Okay, um, hi, I'm in a online she-school. Um, I want to ask him, how does energy equal matter squared?

• 19:38Speaker 04

  Ah, you're referring to my most famous equation, E equals mc squared. 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 are 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?

• 20:23Speaker 04

  For example, consider the sun our primary source of energy.

• 20:28Speaker 03

  Um, because it can also turn into mass because, um, I think that the thing the sun is made of, plasma, is a form of matter and, um, is matter, yes, matter is mass.

• 20:47Aria Finger?

  So I typed that into Khanmigo and let's see what they responded.

• 20:51Speaker 04

  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 equals mc squared.

• 21:17Aria Finger?

  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 through 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 going to change?

• 21:52Speaker 00

  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-age students, high school, the world is using these tools. And so I think it is healthy

• 22:35Speaker 00

  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, in a business class in college, they might have said, hey, I want you to write a business plan. Now, the professors say, I want you to start a business, and then use, and use whatever tools you need to write the business plan. But it's, it's going to be your plan. So it's not like you can just check out and, 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, at Khan Academy, especially with Khanmigo, is, I would say, a middle path that is actually

• 23:14Speaker 00

  probably going to be the, I think, the dominant path. So we, 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, 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, in the same way. And what we're 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 Conmigo. But a teacher can do that. And then, and then through the AI, make the assignment to the students.

• 23:56Speaker 00

  So when the student goes onto the platform, Conmigo tells them, hey, we got to write this paper on the following topic. And the teacher can decide how much support that the AI can give, but it's not going to do the essay for the students. It's going to work like a ethical, thoughtful writing coach alongside the student. And they might riff a little bit on what a good thesis statement 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, if we were collaborating on a Google Doc, you would highlight, hey, so I don't know if this wording is exactly right. 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

• 24:40Speaker 00

  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. It might've been our friend. Even before generative AI, if you do a, a search for, um, 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, lowers the cost of that. But what, what happens now is that the AI, Conmigo, when it reports back to the teacher, it's not just going to report the output. It's going to say, hey, uh, myself and Aria, we worked on this for about four hours. Uh, she had a little bit of trouble initially with the thesis statement, but we got there.

• 25:25Speaker 00

  I had to, you know, some of the resources, 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 worked together. So I feel pretty good. Now, if on the other hand, Reed goes out there, uses ChatGPT or uses one of these essay mills and copies and paste that into Conmigo, Conmigo is going to tell the teacher, I didn't work on this. This is shady. It just showed up. And, and so not only will AI support students better and, and give them better faster cycle feedback, which you need for anything to be better, but I actually think it undermines decades, if not centuries of, of the potential for, for students to cheat in that way.

• 26:07Reid Hoffman?

  One thing that you already covered a little bit of, but I think is 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, wow, I don't want to change the way that we've been doing it for the last 50, 80 years. And you're like, no, no, no. A, you're gonna have to, but B, it's going to be magical.

• 26:35Speaker 00

  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 might be more excited about. Because the pitch to teachers, Khan Academy's been making this pitch and pretty much every ed tech 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 want to address their learning gaps, but they're already overworked and you're giving them one more tool to learn, one more thing to implement.

• 27:17Speaker 00

  And so it's hard and it's, it's present even thinner. What's powerful about generative AI, we still are going to be able to make those cases and it's early days where 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 going to advocate that you should somehow take the teacher out of the loop because it's not perfect. The teacher should be there, but the generative AI can make it a lot easier for them to tweak the lesson plans.

• 28:01Speaker 00

  To say, hey, based on the, well, let's construct the rubric together and based on it, here's a first pass on what I, Konmigo, think the student scored. But you, the teacher, need a, but if we can take that 15 hours a week and make it an hour and a half a week. That's a massive, hopeful gift to teachers 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, 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. And we have a lab school out here, but we're also talking about places like Newark, New Jersey. We're talking about Hobart, Indiana. And the teachers are telling us, one, they love the teaching assistant aspect of it.

• 28:46Speaker 00

  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, aren't bringing them up. And they're, and they're able to save time for themselves and feel like they're supporting their students better. So it, it does feel, read to your point, like it is magical and it's in a, 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, uh, which I think will make it cooler.

• 29:15Reid Hoffman?

  The other thing, one is the teacher, people don't realize it is be great for the teachers. Teachers can spend more time with students, they can spend more time on less, less on roadwork and more on, you know, helping with, you know, part of what the great teachers get into this is they love to help children, you know, and students. It's like more time doing that. The other one is also, um, which we, we hinted at a little bit, which is, you know, wealthy families can always, of course, afford tutors, afford other kinds of things. Uh, I know it's very early days, but have you seen any of the kind of con Mego expanding that aperture more? To the various, you know, kind of more impoverished communities and so forth that can go, now I get a tutor too.

• 29:56Reid Hoffman?

  Um, and that kind of that, that, that elevation of talent, that, that, that, that inclusion in, in great outcomes. And if, if there's anything in particular, please share.

• 30:06Speaker 00

  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-Khan Migo, November of last year. And it's actually turning into, you're going to hear more about the North Ward of Newark, New Jersey. They have some amazing leadership at the ward level and is now expanding to the district level. So when Khan Migo 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. 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 friendly, you know, and I feel really supported.

• 30:48Speaker 00

  And Khan Migo 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 it. In fact, in that way, it's better than even a $50 an hour or $100 an hour tutor, because they're in the middle of the night. They're 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're going to make Khan Migo likely transcend Khan Academy. So it's not just when you're doing your Khan Academy work. It can be your, it can be your, your guide when you're doing anything. But, but also, you know, proactively make sure you're engaged. So right now it's helping you when you're stuck.

• 31:31Speaker 00

  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 talk 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 a Khan Migo 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 Ari is 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.

• 32:15Speaker 00

  But, but I think, yeah, and it's, it's amazing how fast the technology is advancing. And it's really just an engineering question of putting the pieces in the right way that feel, to your point, read magical. But I think that's, that's going to happen in the next three to five years.

• 32:30Aria Finger?

  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, analog ways. And there's so much research that says if a, you know, 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. And so again, the learning is so critical, but also what it opens up teachers to be, you know, these, you know, real life companions that get kids and know them and sort of ask them the right questions.

• 33:13Aria Finger?

  So one question I had is that, you know, as many parents during COVID found out that remote learning was magical and terrible and, you know, screen time, a lot of, you know, 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, Khan Migo, et cetera, when it is sort of intermediated through a screen?

• 33:38Speaker 00

  Yeah. And, you know, I've said this before the pandemic, and I believe it even stronger during or after the pandemic, is 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, hey, it's not about lecture. You can have teacher-directed dialogue or simulations, but ideally kids are working with each other and, and it's, when humans are together, they should be interacting.

• 34:20Speaker 00

  But we know that the status quo in most classrooms are still, kids sit passively, quietly, they're not allowed to talk. But you know, even if your friend says, hey, do you know how to do problem number three? And you say, oh yeah, it's really easy. You just have to remember, you know, someone goes, shh, shh, you can't talk. I'm thinking about, they're shutting down an interaction that's incredibly powerful. Like one student's explaining their understanding and other students 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. And, 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.

• 35:03Speaker 00

  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. 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.

• 35:45Speaker 00

  Because when you're doing mastery-based learning, it's not me against you and a curve, and it's not me against the teacher either. Is that person going to 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 want to do that on purpose, to show that there is a way to do this. Now, ideally, those students, even at Khan World School, have in-person context, they could be in a pod, et cetera. But you can imagine, if you're in rural Alaska, or someplace where you don't have access to a really rigorous school, this gives you that.

• 36:27Speaker 00

  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, where are you achieving? And if you talk to these students, they feel very engaged, very seen, and very, very human. So there's ways to do it.

• 37:01Aria Finger?

  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 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?

• 37:22Speaker 00

  I don't think we're behind yet. And if, you know, if we have anything to do with it, we want to 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 2 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 10 years, I hope we're talking in the tens of millions. And now we're talking about a good chunk of of of the US. So 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. 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

• 38:07Speaker 00

  want to be taking the PISA exam. There's a lot of kids in farms in in Western China who have not seen a school in years. So and and would do very badly on on the PISA exam. So so there's that. I think and 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 in the PISA rankings. It would be right up there with Finland and Singapore, et cetera. And it's about the same size as as Finland and Singapore, right? It's actually a fair it's actually a fair comparison. But the the really interesting thing is we have delegations coming to Silicon Valley from Singapore, from China, from Finland, literally. And they they'll always stop by our offices while they're touring Silicon Valley.

• 38:52Speaker 00

  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 when it comes to test scores. But it feels like all the innovation is happening here. And and and my point of view, and you know, this is my American exceptionalist side of me, I'm like, yeah.

• 39:15Speaker 00

  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 want to 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, famously the Prussian education system. It came out of 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. And what I've argued is we should have a more American school system. What does that mean? Well, America, I was born here and I come from an immigrant family. You know, my mom,

• 39:56Speaker 00

  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. We grew up, we have a very good family. Your grandfather was this, this and that. People are going to think. And but to me, that was always a side 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. 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, Reed's one of those people, he meets them

• 40:38Speaker 00

  all the time. Like, you take a risk and you fail, as long as you failed well, 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. 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.

• 41:21Speaker 00

  And so, 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 got to 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 like, you get multiple shots on goal and ideally use more of your class time for, for innovation, for entrepreneurship, for challenging the norm, not, you know, in some, in some cultures, and I think, I think the US actually could learn a little bit of this. I don't, I, I think we've gone

• 41:59Speaker 00

  too far, not revering educators. I think in, 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, this country special and, and bring that more into the education system and then spread it globally.

• 42:34Reid Hoffman?

  Yep. Disruptive innovation, ask questions, and it's a team sport we do together to improve in the world. We're going to see rapid progress from Khan Migo. What are some of the things that, 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, to our educational journeys?

• 43:07Speaker 00

  Yeah. And every time I, 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 vein. There's some obvious things. You're going to be able to converse with AIs like Khan Migo. I think in the next two or three years, you're going to 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? I'm not, I'm not, you know, I meditate a lot and a lot, the, the, 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. And when you start to do that, you realize that your thoughts are very much like a large language model.

• 43:50Speaker 00

  It's, it's kind of a, a series of, of, of competing large language models that are all feeding off of each other. And then you are this awareness that is like along for the ride on, on, on, on, on some level. So, so I think, I think you're going to see.

• 44:07Speaker 00

  This, this become, I'll give, I'll give more, you know, we talked about the writing example is going to be used more for feedback, more for teaching assistance. I think you're going to see. 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 on the application layer, who knows if future models, they somehow put it into the base model somehow, but the. You know, this notion of memory where you're starting to realize that, you know, we're, we're building things that feel a lot like. What we think like dreams are, or what, um, you know, uh, why we sleep, why you're trying to consolidate your memories from the day. Cause you can't remember everything and you start to realize that, oh, you have to consolidate memories. One way, if you want to 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.

• 44:53Speaker 00

  If you want to be able to refer back to a conversation that you had two weeks ago as, as an intelligent AI. So that I, I, I find from an engineering point of view, it's going to make it super cool and magical, but even from a who are we point of view. It's really interesting. I think you're going to see.

• 45:10Speaker 00

  You know, these AIs done well, I know this can sound a little creepy, but, but done well, they're going to 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 multi-year 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, like, tell me more. Are they also doing, you know, are they also struggling in math? Cause right now their writing is a little bit. And the AI can talk about that, like what's, what's, what's their passion? So I'm super excited about that aspect.

• 45:55Speaker 00

  And, you know, I, I, I think you're going to see, uh, as I mentioned, conmigo 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 Go Guardian and them 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 going to 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 if, 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, as we know, this, this, the stuff can code.

• 46:40Speaker 00

  It can, in some ways, it can be a guardian angel on, you know, the, the, the variation that is our internet, that, 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, your ex-boyfriend or ex-girlfriend on social media for the last half hour. How do you feel about yourself right now? Maybe we want to get back to what, you know, what, what you were here to do. I think we could all benefit. I could also be a therapist. I think we could all benefit from something that has our best interests. As, 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

• 47:25Speaker 00

  mental health and happiness.

• 47:35Reid Hoffman?

  We're now going to go to our rapid virus section. Is there a movie, song or book that fills you with optimism for the future?

• 47:42Speaker 00

  You know, I, I just re-watched Contact with my kids last night and I was like, that's a great movie. There's, 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 is what if we can help bring about something that will make today look like a dark age? Cause we will be in such a better place.

• 48:01Aria Finger?

  I love it. So is there a question that you wish people would ask you more often?

• 48:07Speaker 00

  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? 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 this?

• 48:26Aria Finger?

  I think we're gonna have to have a whole second episode.

• 48:28Reid Hoffman?

  Yes, exactly. I would love to ask you those questions in depth, so we'll have to come back to that.

• 48:35Reid Hoffman?

  Where do you see progress or momentum outside of your industry that inspires you?

• 48:41Speaker 00

  I'm, you know, and look, this might be a little cliche. I'm, I'm reading Walter Isaacson's book about Elon Musk, and I know, you know, y'all, and we both know him, and, and, and I know there, there might be dimensions of, of him that not everyone agrees with, but the, um, it, it, I find it very inspiring to see, almost through act of will, um, you know, industries that I thought no one was going to be able to touch, or, or, you know, once again, in our lifetimes, are, are already, you know, whether it's electrification of cars, or space exploration, or interfacing between, you know, the mind and the digital world. Um, I, yeah, it inspired, when I see things like that, I know, he's not the only person

• 49:26Speaker 00

  doing it, other people are as well, but, um, yeah, it, it inspires me to say like, okay, let's, let's not play small ball here, let's, let's think big.

• 49:35Aria Finger?

  Well, on that note, uh, 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?

• 49:47Speaker 00

  If everything breaks our way, um, you know, I'll start, I'll start quote, small on the education side. Um, I, I think in 15 years, 20, like, you, 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, I've been saying, you know, Kaka's mission is free world-class education for anyone, anywhere. If you asked me even three, four years ago and said, well, how are you going to give every child rich Socratic dialogue? How are you going to give every child the ability to to role play in a simulation or to explain or explore ancient Rome? I was like, yeah, I'm not sure exactly how to do that right now, but Khan Academy is going to be around well beyond me.

• 50:28Speaker 00

  So one day you're going to have AI and you're going to be in an immersive world and you're going to have a link straight into your brain. You're going to have these lucid dreams where you're in Rome, whatever. That's that's not that's likely to happen in the next 15, 15 years, and it's likely to be very accessible. So I'm excited about that. I'm excited about AI actually helping potentially to facilitate conversations between people and be that that trusted moderator in certain cases. I think a lot of the rough edges of the internet.

• 51:01Speaker 00

  AI could actually, as opposed to, obviously can amplify some of them with fake news and misinformation and 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 a a 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. But I think if everything breaks our way, their AI will actually allow us to even undermine that. And, you know, the misinformation campaigns can go both ways.

• 51:46Speaker 00

  You can have, you know, some of, you can create enough distrust in in that, you know, dictator's close. I don't know if that's the line of 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.

• 52:06Speaker 00

  And I'll say one last thing, I think if everything goes our way.

• 52:10Speaker 00

  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, 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. And that's not a, that I consider that dystopian because people don't want to hand out. 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

• 52:54Speaker 00

  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 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 network? I think there's a utopian world where in 20 years, everyone does. It's just common knowledge. And then things are going to be pretty good.

• 53:29Aria Finger?

  I love it. Sal, thank you so much for being here today. It was, it was a delight.

• 53:34Reid Hoffman?

  That awesome. And Sal, as always, could easily talk for another few hours.

• 53:38Speaker 00

  Likewise. Thanks for having me.

• 53:45Reid Hoffman?

  Possible is produced by Wondery Media Network, hosted by Ari Finger and me, Reid Hoffman. Our showrunner is Sean Young. Possible is produced by Edie Allard, Sarah Shleid, and Paloma Moreno Jimenez. Jenny Kaplan is our executive producer and ender.

• 54:03Aria Finger?

  Special thanks to Katie Sanders, Surya Yalamanchili, Saida Sapieva, Ian Ellis, Greg Beato, Ben Ralas, Stacy Olson, and Little Monster Media Company.