A groundbreaking change has struck academia, and its reverberations may be felt for years to come. One of Stanford’s first full courses to ever be openly made available online has gone viral. In a matter of weeks it has signed up more than 100,000 students from around the world! Even as I wrote this article, another 5000 joined! As news of the course continues to spread, the ultimate size of the class could reach greater epic proportions – we could easily see interest skyrocket to 200,000 or even 300,000 or more. Classes of 1 million or tens of millions may be in our future. If Stanford can succeed in teaching classes of 100k+ students at a time, what will it mean for education in general?
The secret to professors and teaching staff handling such an unheard of number of attendees lies in the subject of the class itself: artificial intelligence. This fall, Stanford professors Peter Norvig and Sebastian Thrun will offer their ever popular Introduction to AI for free on the internet. Norvig and Thrun, leaders in artificial intelligence, will be using automated systems to help them at every level of the massive endeavor, from deciding which questions to answer to grading final exams. Anyone can sign up, watch lectures, have their homework graded, and take the exams. Everyone who passes will receive a certificate verifying their completion of the course and marking how they ranked compared to others in the class, including the Stanford students who’ll be attending in person. Check out the videos from Norvig and Thrun below to see what the course will be all about.
Attracting 100,000+ students from all over the globe is no simple matter. It helps when the professors are some of the biggest names in the field. Peter Norvig is the Director of Research at Google, the former senior computer scientist at NASA and literally wrote the textbook on AI (Artificial Intelligence: A Modern Approach, with co-author Stuart Russell). Sebastian Thrun is one of the most brilliant minds in AI and robotics, heading the Stanford team that won the DARPA challenge in 2005, and is the leading force behind Google’s Automated Car, which recently drove more than 140,000 miles without human assistance. These two men are giants in their fields, and their Introduction to AI class at Stanford has rightfully attracted large crowds in the past – 200 students or so.
For their inaugural online edition of the course they want 200,000.
Here’s Thrun pitching the course in the first announcement video released mid-July. I’ve included a more recent update released four days ago when they had just 56,000 students enrolled! Yes, nearly 50,000 more students joined in less than a work week.
If Norvig and Thrun’s expertise makes signing up for this course enticing, it’s also the only thing that makes actually teaching the course possible. There’s simply no way two human beings (or even twenty if you gave them a large staff of assistants) could manage 100,000+ students. Instead, a company founded by Thrun, Know Labs, will help automate every stage of the class. Questions will be submitted and then ranked to see which will be answered in class. Exercises will be available to help teach concepts online. Homework will be submitted and likely graded via AI systems. So too, for the midterm and final exams. It’s a class that teaches artificial intelligence and then uses artificial intelligence to measure how well it is teaching. Recursively brilliant. Norvig describes the mechanics of the class in slightly greater detail in the following video:
The online launch of Introduction to Artificial Intelligence has garnered attention from major media sources like the New York Times and the Wall Street Journal, and has lit the blogosphere afire with enthusiasm and anticipation. Yet, what’s often lost in the coverage is the precedent for this success. In 2008, Andrew Ng, the Director of the Stanford Artificial Intelligence Lab started offering videos for his Introduction to Machine Learning on YouTube, regularly garnering 200,000+ views. This outreach was part of a larger “Stanford Engineering Everywhere ” effort which has racked up more than a total 1 million viewers of its course materials online. Ng’s Introduction to Machine Learning will also be available as an online class this fall, along with Introduction to Databases, taught by Jennifer Widom (chair of Stanford’s Comp. Sci. Department). Catch Ng’s pitch for his class, featuring some excellent examples of robot applications for machine learning, in the video below:
A sample of Ng’s lectures can be found here on YouTube.
Introduction to Machine Learning and Introduction to Databases have attracted 20,000+ students on their own. While that’s not nearly the same crowd as Thrun and Norvig’s, these classes demonstrate that the Introduction to AI online course isn’t some isolated phenomenon. It’s a careful step towards a bigger goal – one that extends even beyond Stanford. As they explained to the New York Times, Thrun and Norvig were inspired by efforts like those of Salman Kahn, whose Khan Academy has expanded from a few rudimentary YouTube videos on math made for his cousins, to an extensive library of 2400+ videos, and 125+ exercises available free to everyone. The Khan Academy demonstrates how video lectures can revolutionize the classroom, inverting the way we teach. Students watch and learn at home, they practice at school. Now, a world class university is bringing the same level of innovation to higher education, showing us that online distribution and automated systems can bring college-level courses to anyone with a steady internet connection. A class of 100,000 or even 200,000 isn’t impossible – it’s inspiring.
If it’s as real as we think it is. The running total on the Intro to AI web page (which currently reads around 102k) is just for those students who sign up to register. It can’t predict how many students will decide to ultimately attend virtual classes and participate in homework. Even if everyone who signs up attends, one has to wonder about the attrition rates over the duration of the course. Even Stanford students regularly drop courses, will virtual students who hit the wall half way through the class have the drive (or community support) to tough it out? It could be that a large chunk of the 100,000+ attendees will leave or fail.
But what if they don’t? What happens then? What open enrollment courses like Introduction to AI are moving us towards isn’t exactly clear. Of the 100,000+ students who will take Introduction to Artificial Intelligence course, only Stanford students will receive college credit. Non Stanford students will receive a certificate which, while ranking their work and asserting their completion of the course, is unlikely to serve as a boost to their credentials. Or perhaps it will, if you were hiring someone, would you be more impressed with their degree from Podunk University in the middle of nowhere if they also had a certified success with luminaries like Thrun and Norvig? If so, Stanford University may be devaluing their product, so to speak. Or rather, changing it.
Highly selective universities like Stanford owe a large part of the eventual success of their students to how rigorously they were filtered before they ever even took a class their freshmen year. If hundreds of thousands of people apply, taking the top 5% virtually guarantees you’ll have some of the brightest minds available. On top of that great beginning, Stanford adds exemplary courses, first hand interactions with professors, and all the other benefits of attending a university in person.
Free online courses like Introduction to AI can’t compete with Stanford’s selectivity, but then again, they don’t want to. Open education simply attempts to bring the same exemplary courses, and the closest thing to first-hand interaction they can find, to everyone who wants it. Such a system will probably never replace the highest echelons of advanced education, but it may make a life-changing difference to those around the world who never had a chance of attending those upper echelons. Poverty, regionalism, and language – how many great minds are lost to these barriers every year? Cheap internet access is expanding to every corner of the globe through mobile devices. Text for all three of Stanford’s pioneering courses are being translated. So too are programs like the Khan Academy. It’s only a matter of time, I think, before online education conquers these great barriers and opens university course work (classes in artificial intelligence no less!) to the world.
In the near term, expect to see more universities joining this bandwagon. Many already place lectures online, and using AI to facilitate actual graded coursework seems like a logical next step. Private companies are likely to get in on the action, too. “Celebrity Lectures” or something of that kind would undoubtedly be very popular (does TED ring a bell?). This won’t be isolated to the US, organizations from all over the world have potential here. I’d be remiss if I didn’t point out that the Indian Institute of Technology has been offering an introductary AI course on YouTube since 2008. These types of classes are going to get bigger. If enrollment is allowed to continue, I’ve no doubt that Thrun and Norvig’s Intro to AI will keep rocketing towards 200,000 students (the count went up by 2000 just while I was writing this article). Powered by online videos and automated grading systems, there’s no reason why these kinds of classes couldn’t reach 1 Million or even 10 Million students at a time! This is a very exciting point in the development of digital education. Major educators at major institutions are seeing the potential in opening their doors. Success now can fuel this trend to keep growing until determination and time are the only barriers to education left.
[image credit: BigCats Lair via Wikicommons (modified), Introduction to Artificial Intelligence (Stanford)]
[video credits: Thrun, Norvig and Ng at Stanford University]
[sources: Introduction to Artificial Intelligence (Stanford)]