Click here to login if you're an NAE Member
Recover Your Account Information
Author: Sal Kahn
RON LATANISION (RML): We are glad to have a chance to talk with you, Sal. I have always felt that education is a means of integrating people into the culture of not only a country but also the world. Given all the things you’ve done over the past 10 or so years, I am very impressed by the evolution of Khan Academy and your involvement.
SAL KHAN: I appreciate that.
RML: As a starting point, I understand that when you left MIT you began working as a financial analyst?
MR. KHAN: Actually my first job was as a product manager at Oracle. My background was in computer science and math, and my first job was in software. Then I left to do a startup where, at the ripe age of 22, I was the CTO—like any self-respecting 22-year-old in 1999. It did well until Nasdaq popped.
Then I applied to business school, where I took a capital markets class with a professor who was an ex-engineer who became a derivatives trader and then decided to get his PhD in finance and became a professor at Harvard. I had a rapport with him and loved his class—I was kind of the closer for the class. I thought, ‘What in life can I do that is essentially this class?’ He said, “You should go work at a hedge fund.” I said, “That sounds great. But what is a hedge fund?” I talked to some people. I had a lot of student debt so I wanted to know how much do you make at a hedge fund? When I found out I said, “Oh, yeah, I am definitely working at a hedge fund.”
RML: I understand you had a cousin who wanted some tutoring in math and you started doing that and that led to what you do now. It’s a pretty big leap from a hedge fund financial analyst to launching an educational enterprise like Khan Academy. How did you get to that point? What was the motivation?
MR. KHAN: The initial motivation was that I have always enjoyed tutoring and math and science and academic knowledge. As a hedge fund analyst I became aware that, no matter what career you choose, there are parts of your brain that you enjoy but don’t get to use. When my 12-year-old cousin was put in remedial math class in 7th grade, I thought, ‘That’s going to follow her through the rest of her life.’ I felt very confident—she is a bright young woman—that with a little help she might be able to do a lot better. I had to help my cousin.
The other motivation was, it’s cool to have a daily connection with your 12-year-old cousin who lives on the other side of the country. I think everyone craves a family connection, especially when you’re by yourself.
And a lot of times the story is Oh, gee whiz, I just happened to fall into this thing. It’s true. Even when I was at MIT, I spent a summer with a fellowship to work on educational software. Many times I had found myself interested in education and the potential for software, in particular, to help scale. I think almost every software engineer has thought about this because most care about education as well as software.
I started tutoring my cousin and in 2–3 months she went from being a remedial student to being the top student in her class. This experience validates that famous Benjamin Bloom study, the Two Sigma,1 about having a personal tutor. Then I started tutoring her younger brothers. Word got around the family, and within a year and a half I was tutoring 10–15 cousins and family friends.
As an engineer I started to think, ‘How does this scale and what parts of this can I create tools for that will help me scale?’ In the beginning my cousins just needed practice. I looked on the Internet, there were some questions here or there, but it was pretty fragmented and of variable quality.
So I started writing questions that were auto-generated and then auto-graded. Then I wanted to see what they were doing so I added dashboarding and tracking. I’m always assigning different questions, so I thought, ‘What if I auto-sign based on what they show proficiency in and what they did not show proficiency in?’ I was falling into this thing.
I was building out a mastery learning system. I did not know that was the name for it at the time, but this is what Benjamin Bloom theorized, and before him there was Carleton Washburne and the Winnetka plan 100 years ago. Many people have already thought of this stuff. I only found out about it when I started doing research for myself.
What really put us on the radar was a friend who asked, “How are you scaling your lessons?” I said, “I’m not.” He said, “Why don’t you make videos of them and upload them onto YouTube?” I told him, “YouTube is for cats playing piano, not serious math.” But I got over the fact that it wasn’t my idea and I did it. (In fact my cousins said they liked me better on YouTube than in person.) The first videos were actually just to supplement the software I was creating.
Because it was on YouTube it was openly accessible, so soon people who were not my cousins were watching it and it started to take off. It did not go viral in the traditional sense, but over the next two, three, four years, tens of thousands of people were coming on each month. This is what eventually led me to set this up as a nonprofit, quit my job, et cetera.
RML: Did your experience either in EECS at MIT in developing software, for example, or your MBA lead you to think there might be a business model that would be workable to deliver something like this?
MR. KHAN: When I was doing it at the beginning, it was just like a significant hobby. When I started I was based in Boston. The firm I was working for was very small—just me and my boss. We moved out to the Bay Area, in the middle of Silicon Valley with all these people who graduated from MIT or Harvard Business School, all entrepreneurs or venture capitalists.
At first a lot of them wanted to know, What is the business model? How are you going to make money? I said I was just doing this for fun. They sounded very irritated. I said I enjoyed it—“look at this nice letter I got, that makes it worth it.” But once there were several hundred thousand people using it, we started trying to get some press. And some venture capitalists started to reach out and say to me, “I will write a check right now, Sal, and you can work on what you want to work on and it will be a for-profit business.”
I think that could have worked. But I had that experience in the late ’90s: everybody is kumbaya when a startup starts and you say, “I will change the world.” When things get tough, or not even when things get tough but after a year or two of the kumbaya, investors will be asking, “Where is my return? Why aren’t you charging those people?” I knew that was not what I wanted.
I had never worked at or run a not-for-profit, but I was intrigued. In my day job I spent a lot of time talking to people and I saw how much capital structure motivates behavior and how your investor base motivates behavior. So I set up a not-for-profit. In a lot of ways this was a bolder bet because as risky as a for-profit venture-backed business is, the data indicate that a lot fewer not-for-profits make it to scale.
RML: In terms of your long-term vision for Khan Academy, you have been operating for close to a decade. Do you see any midcourse corrections? a long-term goal that would lead you to change the orientation? How do you see the whole enterprise going forward?
CAMERON FLETCHER (CHF): How will it be sustainable?
MR. KHAN: At a meta-level, with this whole notion of being a not-for-profit, our mission is a free, world-class education for anyone, anywhere. Every day that goes by I think, ‘Wow, that was a really good decision,’ and it keeps getting reinforced.
There is a Harvard Business School case study on Khan Academy, where the central question is, Should it be for-profit or not-for-profit? The main arguments against not-for-profit are: Will they have access to capital? Will they have access to talent, in order to tackle this huge, free, world-class education for anyone anywhere? I wasn’t even sure about it when it started.
But now the simple answer for both of those questions seems to be ‘yes.’ Not only are we getting decent talent, we are getting the best talent in Silicon Valley. Engineers and designers are hot commodities in Silicon Valley: everyone we give a job offer to has offers from Google, Facebook, Apple—and we are getting 70 percent of them. We have a higher yield than any of those other companies.
We pay well—our cash compensation is close to what people might be able to get from some of these other companies—but we don’t give stock. So the total compensation is maybe 50–60 percent of what you would get at some of these competitors. But it shows that if you pay people enough and give them a cool mission and intellectually challenging work, they feel like they are not a cog, like their work matters.
Some of the top people in the field are on our team and some of them are not driven by money at all. Craig Silverstein was Google’s first engineer; now he runs our infrastructure. He obviously could do anything with his time, but he wants to do this. We have people out of college, too, who could do anything and they are coming here.
The other thing is access to capital. I quit my job on the premise of the social return on investment here. I thought, ‘If I were a billionaire I would give to this.’ That may be a somewhat naïve point of view, and it took a little time, but now our funders are disproportionately engineers. They are people who care about education. They understand the scale we have and the types of measurement and tooling that we are capable of. They appreciate our analytics and how we try to measure and quantify what we do.
RML: In terms of sustainability, when you think about a university, for example, endowment is a very important part of the mix of financial resources at places like MIT and Harvard and other institutions. Are your donors contributing in the sense of operating funds or do you put some fraction of the resources into endowment? How do you envision sustaining the effort in the longer term?
MR. KHAN: Right now our revenue makeup is about 40 percent of our budget, which in 2017 is $30–$35 million. About 40 percent of that will come from some of the folks I just talked about. The Gates Foundation and other large foundations are giving $1–3 million a year.
Another 40 percent of our revenue is from what we call “partner revenue.” It is not pure philanthropy because the partner is getting something out of it. For example, Bank of America sponsored financial literacy, and they get to tell the world, “We are sponsoring financial literacy on Khan Academy”—it’s part of their marketing that they care about education. The College Board gives us revenue to provide free practice for the SAT, which supports what they are doing and is great for us because it super aligns with our mission. We are talking to other testing companies.
We are finding that we can have a revenue model that in no way violates our mission. For example, SAT practice is free for students—it is a key part of their education—but it’s in another party’s interest to make it happen, so that party gives us support.
Another 20–25 percent of our revenue comes from small user donations. We have 300,000 people who give us on average $20.
Hopefully we will always be able to make the case to large foundations that this is a very high social ROI and they will always be part of the mix. But I think the small donors may be a more predictable stream and we can do a lot more there. Right now it’s about $6–7 million a year; I think it could be $10, $15, $20 million a year if we make a little more effort. As our impact scales, I think that will happen.
We are working with more folks like the College Board, doing corporate sponsorship of content. We are a completely noncommercial site—you won’t see any ads. But Amgen, as a real example, cares about biology education and would like to be associated with educating the next generation of biologists and researchers; they give us several million dollars as our biology sponsor. Similar to Bank of America.
But there is a very clear line. None of these donors have any editorial impact. It’s like the Olympics: “We are a proud sponsor of algebra and Khan Academy,” but it is noncommercial.
RML: I get it. You mentioned involvement with MIT. Does that involve their OpenCourseWare activities? What sort of interaction do you have with people there?
MR. KHAN: I don’t have any formal interaction with MIT, but obviously I interact a lot with it. I just saw [MIT president] Rafael Reif a couple of weeks ago at a conference. We always have fun conversations.
RML: Do you have any interaction with folks in the Media Lab?
MR. KHAN: I know a bunch of folks at the Media Lab, but we don’t have any formal partnerships with them.
I do always cite OpenCourseWare as something of an inspiration for me. It reinforced my desire to make Khan Academy not-for-profit.
I remember in the late ’90s there were a bunch of for-profit startups partnering with colleges, and colleges were themselves saying, “Hey, this is a revenue stream. Let’s make some money off this.” Then I remember seeing the press release that MIT was giving it all away for free. They were putting money into it on principle. I found that to be incredibly inspiring. I felt very proud to come from MIT.
A lot of colleges talk about teaching ethics, but to take such a bold stance at the time, that is probably the best ethics of all. I gave a commencement address at MIT in 2012, and said this was an inspiration for me and that people should be proud.
CHF: You may be interested to know that the fall 2016 issue of The Bridge was devoted to open educational resources, starting with MIT.2 And this interview will be published in the spring 2017 issue, which will be on ethics, so you have that to look forward to as well.
RML: Talk about connectivity.
CHF: Going back to the matter of scaling, I see that you started off in the sciences and math and you have expanded to the arts and humanities. How do you pick the subjects that you are going to expand into? And are you considering other topics in those or other areas?
MR. KHAN: Yes, we look at where there is the most need and most demand. We are pretty comprehensive now from kindergarten through the core of college math. In another month or two, we will be almost fully comprehensive in introductory chemistry, biology, and physics. We are going into world history, American history, civics and government. So we are definitely expanding.
For us the vision that I have articulated with our team is that in the next 10 years we should cover every K–14 core academic subject—all the reading, writing, math, social sciences, the sciences, from pre-K or elementary school all the way through the first two years of college.
CHF: That is an unbelievably expansive ambit. I am wondering what are the pedagogical guidelines? How do you ensure that the team members you bring in have the right expertise to design content for such a range of courses and grade levels?
MR. KHAN: That is a central question. For K–12 math, for example, the Common Core standards set up at least an architecture. We have four or five in-house folks on it, then we work with teachers and researchers, and start writing items and creating content. I still create a lot of the videos but for math it would be vetted by Bill McCallum’s group at Illustrative Mathematics. Bill McCallum was one of the two authors of the math Common Core standards. They didn’t do this for anyone else.
This is kind of a secret weapon of being not-for-profit. They were not willing to do this for McGraw-Hill or Pearson, but they vetted every item, every piece of our content, to ensure that it is consistent with the spirit of the Common Core. This is a big deal because you can give the Common Core standards to a teacher, but how it gets implemented, who knows. With us, they can validate the words used, the visualizations, to make sure they’re consistent.
As we got into higher-level math and calculus we developed that content in conjunction with the math department at Phillips Academy Andover, which is where the Advanced Placement (AP) test started. They are one of those places where 150 kids take the test and 149 get a 5 on it.
We developed content with those teachers and now we are working with the College Board to vet it and make sure it’s exactly what they have in mind. They really get excited because, before, they just created the test and created some standards, and now they can say, “Hey, this is a pretty good reference curriculum.” Teachers can point their students to it. If a student goes to a school that doesn’t have AP classes, they have an alternative now in Khan Academy.
As we go into world history, civics, and government, we are talking closely to folks like the College Board.
In terms of the people we hire, let’s say for chemistry: our chemistry fellow is a PhD in chemistry from MIT and has taught all of the core chemistry classes. In addition, we usually ask them to create writing and even video samples, to see their take.
It is one thing to cover the standards—textbooks have been doing that forever, but they have also been really dry and hard to digest and they are not clear. So you can cover the standards, and even be pedagogically correct, but I would say the harder dimension is to really connect with people. Do they really feel comfortable with the material? And does it connect the conceptual dots? That’s the harder thing that we are getting better and better at looking for.
CHF: Especially when you branch out beyond the College Board or the SAT to the lower grades. If you’re talking about math or art or history for pre-K and first grade, for example, it seems to me that those would have different pedagogical needs. Do you have external reviewers?
MR. KHAN: Yes. And for the early learning stuff another exciting thing has happened recently. There is an app company called Duck Duck Moose—I have always cited it as best in class. It was started by a husband and wife who did it for their kids, so it is kind of a parallel narrative to what I was doing. They reached out to us a year ago and said they wanted to become a part of Khan Academy and were willing to donate their organization. They became part of Khan Academy and all 23 existing apps are now free. You can download them right now and start using them. Duck Duck Moose has experts on early learning, but they are also working with the education school at Stanford and other places.
RML: How many folks do you actually have? I was going to say subscribers. That may not be the right word. How many people do you reach a year?
MR. KHAN: There are different numbers. Some of these you could put in the category of vanity metrics, but more than 50 million registered users total, and 10 million people take a learning action per month—i.e., they did at least one thing.
CHF: Does that mean they completed a unit?
MR. KHAN: Yes, a unit for us could be as little as a video.
CHF: I was wondering because I went on your site and it says there are about 50 million learners. How are those measured? Is it the fact that I, for example, signed in so you guys now have my email address—does that make me one of the 50 million?
MR. KHAN: Yes, I think that’s the 50 million number you are seeing. That grows by 2–3 million people per month. If you look at unique users over a year, that number approaches 100 million. And the monthly number of people who take a learning action is 10–11 million right now.
RML: Do you know anything about the demographics of this population? How many of these are people who are beyond the K–12 system? The intention is that anyone who wants to learn something would have access.
MR. KHAN: We are trying to get better data. The bulk of our users are in the middle school through early college age range. We do have a lot of adult learners. It’s hard to get precise information because a lot of our users are not registered and for the registered ones we don’t ask a lot about them. I would estimate probably 30 percent are out of school and doing it out of interest or trying to brush up to help their kids. And about 35 percent of our usage is outside the United States.
CHF: You do have an impressive array of languages into which your materials are translated.
MR. KHAN: Yes, that is exciting, and it’s growing. We have a full offering in five languages—the software is fully translated into Spanish, Brazilian Portuguese, Hindi, French, Turkish—and there are 20-some, maybe even 30 languages where we have a reasonable translation effort under way, done by volunteers. There are at least several hundred videos translated. I think there are 10,000 videos total in the entire library right now.
The other dimension of the vision is every core academic subject from kindergarten through core of college and in every major world language.
RML: That is stunning.
CHF: When you say that 30 percent of your users are older—defined as anybody beyond college—how much of your resources are used for continuing education and do you make any effort to reach out and do some promotion for that use?
MR. KHAN: Our number one testimonial comes from the continuing education person, the person who wants to go back to school and feels intimidated, especially by math. But we have to do a much better job about communicating it, tying it more closely to opportunities.
CHF: What do you see as next steps for yourself?
MR. KHAN: We have got our work cut out for us. We want to make the platform much more engaging. We see opportunities. We already have tools for teachers; how can we help them even more to free up time for more interactive things with their students? More one-on-one time, more projects, dialogue, things like that.
We hope to tie learning on Khan Academy to actual opportunities. This is already true for some things like the SAT. But maybe you’re an adult learner and you prove yourself on Khan Academy and can take the actuarial exam now. If you do that you can have a pretty good career. Based on this work, we think we can give you a nurse apprenticeship or something like that.
That is not going to be us by ourselves. We are never going to be an assessment body or do job placement, but we can work with the people who do to create clear connections between learning on Khan Academy and some type of opportunity.
RML: Yours is quite a remarkable venture and a laudable effort. I am particularly impressed by your view of engaging people and educating them on a worldwide scale. That is important not only in terms of careers and personal well-being but also in terms of bringing people together around an educational goal that may help support world harmony.
MR. KHAN: I appreciate that and I agree with you.
RML: Before we close, is there any message you want to convey to the NAE members and other subscribers to The Bridge, who include members of Congress and deans of engineering throughout the country?
MR. KHAN: I think there is a lot of pessimism about what is going on in education. In the United States there is concern about a globalized world with automation—a world with a lot of productivity, but who is participating in that productivity?
As we transitioned from an industrial model, which is a bit of a pyramid, we had a lot of labor at the bottom and in the middle class, in information-processing jobs. A lot of labor in the industrial society was pretty good middle-class jobs, factory jobs, and the like. At the top were the owners of capital, the creative class, the researchers, artists, people like that.
Now the writing is on the wall. First through globalization, and shortly thereafter through automation: not only the bottom layer, through robotics, but even the middle layer, around information processing, is going to be automated. We are going to be a more productive society, but where does all that wealth go? Does it go to just the small top of the pyramid? Do we have an unstable society? Do we have to do some type of aggressive redistribution to have a stable society?
My hope, and I actually think it is possible, is that we invert that pyramid where instead of 5 or 10 percent of people at the top of the pyramid, maybe we can get 40, 50, even 60 percent of the population there. I call that the “Star Trek Reality,” where pretty much everyone is a researcher, an explorer, an entrepreneur, an artist of some kind.
CHF: What a beautiful optimistic view.
RML: Yes, it is. I hope that view becomes reality at some point. It would certainly be an asset to the whole planet.
MR. KHAN: We have 400 years to make Star Trek come true. I am working on it.
1 Bloom BS. 1984. The 2 Sigma problem: The search for methods of group instruction as effective as one-to-one tutoring. Educational Researcher, June/July. Available at http://web.mit.edu/5.95/readings/bloom-two-sigma.pdf.
2 The fall 2016 issue of The Bridge, with the story of MIT’s launching of OpenCourseWare, is available at https://www.nae.edu/Publications/Bridge/162252.aspx.