Coursera co-founder Andrew Ng is widely considered a pioneer in the field of artificial intelligence. Along with Daphne Koller, he is the co-founder of Coursera, the massive open online course (MOOC) platform, in April 2012. In just a little more than three years, Coursera has over 12 million users enrolled in more than 1,000 courses from more than a hundred institutions worldwide. Ng taught at Stanford University and is the director of the Stanford Artificial Intelligence Lab. He works on deep learning algorithms, which Ng says are loosely inspired by how the brain learns. He worked on one of the most ambitious artificial intelligence systems at Google called Google Brain. The system analyzed millions of photos taken from YouTube videos and learned to recognize objects, including human and cat faces, without additional human guidance.
Last year, Ng announced he would be stepping away from his day-to-day responsibilities at Coursera to become chief scientist and head of Baidu Research, and lead a new five-year, $300 million research initiative based in Silicon Valley. Chinese-language search engine Baidu, sometimes known as “China’s Google,” is the world’s fifth-most popular website with a $55 billion market capitalization.
In an interview about MOOCs and their impact, Ng says they allow universities to take their great content and project it onto a larger audience than they ever did before. A recent study co-authored by Wharton professor Ezekiel J. Emanuel on the impact of MOOCs on traditional business education, also found that rather than poaching students, MOOCs complement, enrich and help business schools reach new diverse audiences.
An edited transcript of the conversation is below.
Knowledge at Wharton: What role have MOOCs played in university education?
Andrew Ng: I think MOOCs allow universities to take their great content and project it onto a larger audience than they ever did before. Perhaps not surprisingly, this allows universities to reach a much larger, much more diverse audience than has ever been possible. I think this knowledge is so radical in everyday society and most people on the planet will never have access to an on campus walk-in class. I kicked off a Coursera founders’ conference by telling the story of one of the students, a baker in Bangladesh. She took Coursera courses — including a microeconomics class from the University of Pennsylvania and model thinking from the University of Michigan — and learned how to run a business. I showed her statements of accomplishments using her verified certificates on the big screen at the Coursera founders’ conference. Here was a woman who could never attend classes in her city, but today she credits part of her success to Coursera courses.
“I’m very proud of the reach that we have, which is vastly greater and more efficient than a traditional on-campus U.S. university.”
Knowledge at Wharton: All over the developing world, people for whom English is not their first language are trying to use online education to not only better their own skills, but also to help catapult their companies’ goals as well. They are very eager to learn just so that they can make things better for themselves, whether it’s a career, or it’s just learning something new, just like the baker in Bangladesh. What do you think about the impact that Coursera has had, and how you think that might change in the next few years to come?
Ng: A lot of the students who come to Coursera today are from developing economies, and it’s true that India and China are just two pieces of it. I absolutely agree that the developing world is much more than just that. The debate on developing economies is an interesting one, because there is the hope for MOOCs to give access to the niches of society.
I’ve actually heard two messages used. One sometimes will criticize that two-thirds of the people we serve are not in developing economies and so we’re serving two-thirds-developed economies. However, in almost in the same breath, a different person encourages us to increase our learning platform for [that segment]. I think that all Coursera was doing was serving the call to the half-million learners that all come from developing economies. I think everyone will agree that’s a fantastic thing. And that is actually what Coursera does today. It’s just that on top of that, we serve another five-and-a-half million learners that come from developed economies. So I think that there will be a trickle-down effect, as more people get Internet access. There are billions of people on this planet who don’t get good Internet access. I’m very proud of the reach that we have, which is vastly greater and more efficient than a traditional on-campus U.S. university. I think the reach of MOOCs has a vastly greater representation in the less privileged members of our society.
I think the language issue is a very interesting one. Sometimes I’ll go to a university, and I’ll ask them, “How many of your professors speak English, French, German, Italian, Chinese, Spanish and Arabic?” I think usually they’ll find none of their professors speak that many languages. I hope that in the future, they will on Coursera. We don’t today and we’re not where we would like to be, but I think that the opportunity to translate content into lots of languages means that professors on Coursera will be able to speak English, Spanish, Italian, German, French, Chinese or Arabic. And I think that having a community of learners and translating content will help, say, English-speaking professors reach a lot of non-English speaking learners. We’re definitely not where we’d like to be and there’s a lot of work to be done still, but I think the early initiations of having a translating community, of having student volunteers translating all the content, seems very promising to me.
Finally, just to relate a kind of personal story: One of the things we’re trying to do is remain connected with learners, and so when I travel to different cities, I will sometimes go to the local student meet up. There are thousands of student meet ups organized all over the world, and when I’m in some countries, sometimes I will reach out to the organizer of the meet up and request to join their group just to meet and chat with students. I do this partly because I enjoy meeting the students and partly because it helps me stay connected to learners all around the world. And when I do this in developing economies, one thing that struck me was the real hunger for knowledge.
In the U.S. when I go to some meet ups, there is a certain hunger for knowledge. When we walk around our university campuses, we are used to information overload, there is just a buffet of great causes, and we just don’t have time to learn everything we could learn. In developing economies, these people really don’t have access. I’m very surprised when people would take a train for two hours to listen to people talk in some city because in the rural towns in which they live they just did not have access to knowledge. When I speak with all these learners about MOOCs, there is a certain gratitude and hunger that’s different than what there is in the developed world.
Knowledge at Wharton: In many studies about developing economies, we have heard about the notion of leapfrogging. Clearly there are examples like ignoring landlines and going straight to cellular phones. How do you think these models are changing in those parts of the world? Given the availability or at least some awareness of the MOOCs, will there be a different model of education that comes up in those parts of the world? Is that possible? Or have you already seen some signs of that?
Ng: Yes, there are some early signs. Countries like the U.S. or China have relatively developed higher educational systems that are serving a reasonably large fraction of the population. India, in contrast, has a very small higher educational system relative to the size of the population. And their big challenge is to create more capacity. The U.S. does not need to create a vastly larger higher educational system. I think we tend to worry more about issues of cost equality. I think it’s very good that we worry about how to make it better. India just does not have enough seats.
And so, some of the most innovative work I’ve seen is transforming the higher ed system, especially in India, where they have incredible programs, creating tens of thousands, and hopefully soon, hundreds of thousands of teachers. They blend the learning fashion, combining MOOC content with closer instruction. I think that blended learning was a great idea. I think it will improve the quality of education and I think that the first few countries to implement this skill will really influence the way that really makes them appear valuable to the whole country. It will be developing economies, specifically India and China.
The challenge for blended learning will be in teaching. If you want to move your country to blended learning, this is something that requires working with hundreds of thousands of teachers to help them understand a better way for them to teach. And I think it’s clearly a better way for them, it would add more value for the teachers, more value for the learners.
Knowledge at Wharton: It’s interesting you bring up India because a relatively large percent of people enrolled in Coursera are from India, much more than from China. The completion model that a lot of people question about for the MOOCs really doesn’t apply to the developing economies, since people there are just using MOOCs to make themselves and their lives better. However, what do you feel about the whole notion of completion rates? And it doesn’t really have to be about the U.S. and maybe Europe, but have you seen any research coming from the developing world about completion rates for students there?
Ng: We have measured completion rates in different countries and yes, there are small differences in completion rates among different countries. Canadians, for example, have a slightly higher completion rate than Americans do. I don’t remember whether Indians complete courses at a higher or lower rate, but I think that the issues of completion are really global. And I think completion is becoming an increasing fear, [but] it’s the wrong way to think about what learners are getting out of MOOCs. You have 5,000 students that complete a course. There’s about an equal number that watches every single video, but doesn’t do the homework. So should we then say that the completion rates are actually almost double what they are? If they’re watching every single video, presumably they’re getting something out of it, but they just chose not to do the homework. That seems like a fine thing to me.
At Coursera, we are starting to experiment with more flexible models of education. Today we’re asking ourselves questions such as, “For this region or instruction, why do we need deadlines?” It turns out that one of the reasons learners don’t complete the course is that, something happens at work, and they miss a deadline, after which it can be difficult for them to finish the class and earn their certificate. They become demoralized, so we’re asking ourselves, “Why are deadlines necessary?” Deadlines are a somewhat primitive way to motivate course completion — if you don’t do the thing on time, you lose points. But in this era of auto grading and peer grading, is it possible to just learn with a small flexibility and let them complete courses at their own pace? And that’s a really good question; I really don’t know the answer to that. It’s possible that without deadlines, maybe no one will complete courses. My instinct is that won’t be the case, but these are the sorts of questions we’re asking ourselves.
“If you want to move your country to blended learning, this is something that requires working with hundreds of thousands of teachers to help them understand a better way for them to teach.”
Knowledge at Wharton: Some Coursera competitors, edX for example, have adopted a non-profit 501(c)(3) corporate structure, while Coursera is a for-profit venture. Does choice of structure change the way decisions are made at Coursera? For example, edX maintains all data following the Family Educational Rights and Privacy Act Regulation (FERPA), which limits some uses of user data in subsequent marketing, while easing data access to researchers. Would Coursera meet resistance from investors if it chose to do that as well?
Ng: I can’t comment on edX since I’m not familiar with the details of their operations. But Coursera is extremely mission driven, and has always been. Even before founding the company, when the team was comprised of four students and me, our No. 1 rule was “Do what’s best for students”, and this remains our touchstone for all of our decisions today (though it got rewritten to “Do what’s best for learners” later). We also take learner privacy very seriously.
Knowledge at Wharton: Is there evidence that employers are valuing signature track certificates in employment decisions?
Ng: We have a flood of anecdotal data from both learners and from employers. I’m personally very confident that MOOC certificates are helping many learners find better employment. Unfortunately, I don’t hard numbers to back that up yet.
Knowledge at Wharton: Following up on your comments regarding the limited value of deadlines, when Coursera first launched there was much discussion about the social dimension to learning and the work done to encourage a sense of a cohort taking a class together, following what educational theorists would call a constructivist approach. Your recent efforts in developing an on-demand service seem designed to support individual study. Is Coursera changing course with regards to the value it places on cohort learning?
Ng: I think that cohort learning is valuable, and so are learner interactions. On-demand is better in some ways for encouraging those interactions. First, in today’s peer grading system, after you submit your homework you might need to wait a week before it gets peer graded. In an on-demand system, once you submit your homework, we could ask the next five students who submit to grade your homework. Depending on the number of active learners, this will usually mean that you get your homework graded more quickly. Also, in today’s cohort system, the forum answers are erased and we start from scratch each time a course re-runs. In an on-demand system, forum answers can be preserved permanently, and be an archival record (which new students can keep adding to), thus making answers easier to find.
Finally, assuming a course has a large number of active learners, this means that at any point you’ll also find ample peers to discuss the course material with (similar to the original cohort model). There are pros and cons to the cohort system, but in terms of social interactions, I think the pros heavily outweigh the cons.
Knowledge at Wharton: As you know, a relatively larger percent of high school students are attempting and completing MOOCs. Can you talk about the kinds of students taking these courses?
“In this era of auto grading and peer grading, is it possible to just learn with a small flexibility and let them complete courses at their own pace?”
Ng: Right now, most learners that are taking MOOCs are actually working adults. Coursera serves a very broad demographic, anything from high school students to retired adults. But the center of gravity for Coursera is working professionals, so median age is about 35, with a bachelor’s degree, and most in their 20s and 30s, also late 40s. The reason for this is that, just as a statistical fact, most of us spend most of our lives as working adults, only a relatively short period of our lives is spent in high school, college, and so that’s part of why most of the learners are working adults.
The second is, it turns out that most college and high school students already have convenient access to education; you just go to college every day. But working adults such as you and me don’t have continued access to education that we need in order to stay current. Even for professors or people who aren’t professors, it’s very inconvenient to hire a babysitter twice a week and go to a night class at a community college. So I think the biggest impact of MOOCs is bringing working adults back into the educational system.
It turns out that, having a room full of 100 people is a very inefficient way to learn. And we know from the data that the retention rate is shockingly low. You remember 20% of the lecture. Reading is another example. Maybe adults have learned to become efficient at using these modes of learning, but I think sitting in a lecture [is] very challenging for young people. I know more about reading, but it turns out that reading is hard for a lot of people, because there’s so much work, there’s so much text, and it’s challenging for people to learn how to process all that and to identify and focus on what’s important. I guess the same thing is true for lectures, that over time we have developed certain habits of processing the lecture, but I think overall, the modern lecture, where the professor just talks in front of 200 people, is not a very efficient way for learners to learn.
Knowledge at Wharton: Switching back to the developing world, clearly, online platforms need huge investments. Coursera and some of these other MOOC platforms have been able to secure these investments in the U.S., but how do you think that would translate into the developing world, even if you leave China and India aside, since there can be capital there. If you talk about different parts of South Asia or even the Middle East and Africa, how do they go about building these platforms for their people in a fashion that works best for them, whether it’s overcoming the language barrier or the technology barrier? How do you think these models can be developed there?
Ng: I think the world is best served by having one or a relatively small number of MOOC platforms, because technology is just scaled that way. I think Samsung and Apple have done a great job, and you can take an iPhone and localize it to different countries and get a local card, but it makes no sense for every country to invent their own cell phone because they think they need a special piece of hardware.
“Overall, the modern lecture, where the professor just talks in front of 200 people, is not a very efficient way for learners to learn.”
I think an example to would be to say that Penn wanted to teach only students in Philadelphia or if Stanford wanted to offer a MOOC that was restricted to learners in California that would be a very strange thing to do. So in the same way, I think the top, say, Chinese universities will want to take content and share it globally, and I think it’s better for universities, it’s better for learners. Learners want to learn from the best universities around the world, not the best university that they’re living two miles away from. I think the economics of having global MOOC platforms makes more sense, and it’s better for universities and learners.
Knowledge at Wharton: Occasionally people in those local geographies are very proud of what they have done and what they’re trying to do, and occasionally will look at the Western markets and say, “You know what, this is not good enough, or this is not the right model for us, we should build something that is localized and more relevant to our people.” This argument applies to different industries, not just education. I think that some of them look at Coursera and look at the other MOOCs and feel like they were built in the U.S., they have had the opportunity to develop, but will simply say, “Well, this will not work here.”
Ng: I’d love to understand the substance behind that because I think that countries should express their culture through their content, not in software. So I’m proud that today, some of the best [Coursera] courses from China are Chinese history, Chinese traditional medicine and I think this is content that Chinese universities are using to express their culture. And your culture is best expressed with content, not with software…. Ultimately, it’s all about learning. I think MOOCs should be all about making the best possible experience for learners, and so the question is, the fate of the students in India or China, do they want to learn from the best universities from around the world, or is there a different experience that they prefer? And so far, I think we have seen that they would like to learn from the Whartons and the Princetons, and the leading universities in the world.
Knowledge at Wharton: What are your thoughts on Wharton professor Ezekiel Emanuel’s study about MOOCs upending the so-called traditional education model, not just business education?
Andrew Ng: I think that to Professor Emanuel’s point, MOOCs can become a great avenue for universities to source new applicants. I think Wharton, Penn and other universities have attributed their rising applications at the undergraduate level in large part to their MOOC activities. Several universities talk about sourcing new students to attend their on campus programs through MOOCs. And I guess, the sensational question, will MOOCs replace professors? And I think that to answer that, think about your favorite professor back when you were in college. I think if a computer plays that question and answer than I think that answer is clearly no, but I think the opportunity of technology is not to replace them; instead, to free up your favorite professor from the more repetitive aspects of teaching the grading, so that that favorite professor of yours can spend more of their time in conversations with future students as they did with you.
And so I think that’s a blended aspect of on-campus education. I think in the short term, the impact of MOOCs will be in continuing knowledge and education to people who don’t have access to our campuses. Over the longer term, I think that there will be a different change, [that] the campuses will use a mix of online content and in-class symmetry.