Uber for Education
Sebastian Thrun, a leading figure in artificial intelligence, hopes his startup Udacity will be a digital university for a new kind of economy.
In 2012, Sebastian Thrun, an expert in automation and artificial intelligence and a former tenured professor at Stanford University, became the cofounder and CEO of Udacity. There he is creating a rival to traditional postsecondary education: a university focused on lifelong learning, in very small portion sizes, on demand and in a mobile format.
Udacity’s recently announced “Android nanodegree” was created with a $4 million investment from Google. It’s a six-to-nine-month program costing students $200 a month and promising to teach them everything about programming for Android, from the fundamentals to building apps. Projects are graded by Udacity’s network of 300 global code reviewers. As he explains to Business Reports senior editor Nanette Byrnes, those contractors are a key part of Thrun’s current vision for Udacity—a departure from the more common MOOC (massive open online course) model that the company initially seemed to be following, in which professors lecture and students do problems independently online.
Teaching students to master a topic, difficult in any classroom, seems to be one of the biggest challenges facing online learning. You say your approach works. How does it?
We very deeply believe, very passionately believe, that learning by doing trumps learning by listening. We believe the ultimate experience that really makes the master is to do something, build something, invent something, design something, code something. At Udacity, we built an Uber-like platform. With Uber any normal person with a car can become a driver, and with Udacity now every person with a computer can become a global code reviewer. And the mechanics for the code reviewer are the same: you’re being paid per code review, and you’re being assessed by your students. Our global code reviewers, on average, out of a five-point possible score, get 4.8 points. They give students back a very insightful and detailed, human-level, expert-level review of their work, typically within two hours, including detailed feedback on coding style, what works, what doesn’t work, and so on. Just like Uber, we’ve made the financials line up. The best-earning global code reviewer makes more than 17,000 bucks a month. I compare this to the typical part-time teacher in the U.S. who teaches at a college—they make about $2,000 a month.
Is this particularly applicable to teaching coding? Can this translate to other fields as effectively?
I want to be careful not to say every field, but in many, many fields, you learn by doing and not by listening. Fields we are not going to engage in would be something like tennis or violin, where your own practice is absolutely fundamental. But in all my life, honestly—be it math, be it finance; I’ve done some medical studies, I recently became a pilot and learned to pilot a plane—in all these studies the mechanism is quite similar: the most effective learning environment is often one where the student gets to practice something under the guidance of someone more experienced and then gets personalized feedback on how they are performing.
This is a digital way, I guess, of re-creating that mentoring or coaching that a student can have with a teacher.
Our data shows that having this dedicated human touch leads to much more profound learning effects. We couldn’t really do this with computers because they are not smart enough yet. You not only have someone to go to when you need help but also someone to hold you accountable, too, and that has a big impact on learning.
You told me these reviewers could be in the U.S., India, Canada, anywhere.
With Uber you have to be in the same city to drive a person around in your car. With Udacity you can be in Chile and grade someone’s work in Lebanon, and the person in Lebanon wouldn’t even know the grader happens to be in Chile. The grader’s ability to stay in the network to make money is directly contingent on student love. So he or she is going to work really, really hard to make their review extremely insightful. The student wants an insightful review; they pay us money to get the most insightful reviews. By giving them the power to review the grader, all the incentives line up.
You’ve spoken often about your desire to democratize education, to reach a broad audience and reach people who don’t have access to this kind of education otherwise, but Udacity is a business. You charge for your classes while many MOOCs are free.
Education should cost money because of the service rendered. So we decided to have two paths. Access to our content basically costs us nothing—it costs us like 50 cents a student, and we decided to just give this away. If you are in central Africa and you really want to get an education, anything that is easily replicable for us we give to you for free. The reason why we can do this without going bankrupt is this kind of freemium business.
You have a successful partnership with the Georgia Institute of Technology, but some of your university tie-ups did not work out, and now you are only adding new programs developed with private industry. Why do you think you couldn’t succeed with universities?
We chose to go to industry because we believe the future of learning is lifelong and not just one-time. We have data from the Department of Labor Statistics that says the average employment in a job in 2002 was 4.6 years. We know that is shrinking. We know people have seven different careers over their lifetime. As a result, we needed to do two things: make the learning unit smaller than the conventional degree and make it fresher than a traditional degree. At the end of the Google Android nanodegree, we are teaching features of Android that you can’t find anywhere else, that just launched in Android. Google is now running a career summit where they are inviting the top students from the Android nanodegree on campus, all expenses paid, to meet with their engineers and their recruiters. We are effectively building an alternative path to employment and an alternative to existing credentials.
Colleges are very focused on the ages from 17 to maybe 24, but people live 70 to 80 years now in many countries. For people in their 30s, their 40s, military people coming back [to civilian work], women raising children who want to reënter the workforce, all huge factors in the workforce—for those people there’s no educational venues that I know of that work in this country. There’s a huge vacuum.
Universities have become more and more exclusive. In fact, they pride themselves on their exclusivity in admissions, and as a result they attract the best professors because they want to work with the best students. There’s a network effect that only works with exclusivity. Industry says the opposite: make it as cheap as possible so we can reach everybody. I think it’s a much better alignment.
Hear more about AI at EmTech MIT.
September 11-14, 2018
MIT Media Lab