Sign In

Communications of the ACM


Stanford Schooling - Gratis!

Stanford professor Daphne Koller

Stanford computer science Professor Daphne Koller opened her online class on probabilistic graphical models, previously available only within Stanford, to the public in January.

Credit: Steve Fyffe / Stanford News Service

Last October, for the very first time, Stanford University made available online—and at no charge—three of its most popular computer science courses, classes that normally require admission to the extremely selective and expensive university.

More than 300,000 students registered for Introduction to Artificial Intelligence, Introduction to Databases, and Machine Learning, and participated in an experiment that could help transform the way online education is delivered.

"Never before has a premier institution offered a real class experience to the general public, one with meaningful interaction and real assessments," says Daphne Koller, a CS professor at Stanford's AI laboratory. "These courses are a much more meaningful and active experience than the static online course materials that have previously been available to students."

While metrics are not available for the three courses, an online Stanford report shows that of the 91,734 students who registered for Introduction to Databases—taught by chair Jennifer Widom—25,859 turned in some homework, and 6,513 received "statements of accomplishment."

No class credits are given to the graduates. Instead, students received a "statement of accomplishment" provided by the instructor, not the university.

Greg Linden, a research associate at the University of California, Berkeley's Haas School of Business, enrolled in the Machine Learning class taught by Associate Professor Andrew Ng.

His assessment?

"The course seemed a bit light," says Linden. "It could have had more depth, more time commitment, more material, and harder programming assignments.... That being said, it was a great start for many and a very well done class, probably my favorite machine learning class I've ever taken. I have to say—even though I should know this material having worked in the field for some time—I learned several new things in the class."

Peter Norvig, director of research at Google, taught the AI class with Sebastian Thrun, a Google Fellow and a Stanford CS professor, having taught only traditional classes in the past. "I think we learned a new set of idioms for interacting with students in a way that is simultaneously impersonal, because the student can't ask questions that will alter what we present, and highly personal, because we present the material in the same way we would in sitting down and doing one-on-one tutoring," he says.

Norvig believes the online format demonstrated three things: Many people are hungry for high-quality CS education; students can learn by watching videos, answering frequent quizzes, and doing associated homework; and students can get the advantage of peer-to-peer interaction through the online discussion forums.

"This doesn't mean traditional college courses will go away," Norvig says, "but it does mean a large percentage of the educational objectives can be met for a price that's about the same as a traditional course, but reaches 100 times more students."

Koller has been teaching her class on probabilistic graphical models online for two years, but only within Stanford. In January, she opened the online class to the public using her pioneering pedagogical model of short video modules, simple quizzes embedded in the video, significant amounts of interactive assessments with immediate feedback, and a question-and-answer forum in which students interact with each other and course staff.

Meanwhile, Massachusetts Institute of Technology (MIT) recently announced MITx, a new, open-source online learning platform it is currently developing. According to an MIT spokesperson, MITx will offer MIT courses presented specifically for online learning, and students who demonstrate mastery of a given course's material can obtain certificate-based credentials at a low cost. The first MITx courses will be offered this spring.

"It appears MIT saw the benefits in our approach and decided to move in that direction," says Koller. "Given our success, I believe more and more classes will be made available in this format. Thus, I think this online format can give rise to a significant change in higher education."

Back to Top


Paul Hyman is a science and technology writer based in Great Neck, NY.

Back to Top


UF1Figure. Stanford CS professor Daphne Koller.

Back to top

©2012 ACM  0001-0782/12/0300  $10.00

Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and full citation on the first page. Copyright for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or fee. Request permission to publish from or fax (212) 869-0481.

The Digital Library is published by the Association for Computing Machinery. Copyright © 2012 ACM, Inc.



i applaud the wide dissemination of knowledge that the internet makes possible, and to the efforts of MIT and Stanford to make their course materials openly available to the general public. but i take major issue with issuing certificates of any kind, and with claims that there is any real judgment being made as to "assessment of achievement". what significant assessment can be done when you have 25,000 students submitting homeworks, or taking exams? if you wish to claim that a computer-gradable exam or homework assignment can accurately measure mastery in databases or AI, i beg to differ. maybe i am old-fashioned, but i have taught mathematics and computer science at universities for 40 years, and have yet to compose an assessment tool that doesn't require REAL WORK on my part both before and after the exam or homework submission. short answer and multiple choice exams just do not do the job in any course worth its salt!


Yeah, 93% attrition counts as a success for a which world?? And for the 7% who completed the course...will any machine learning company hire them to develop there software?

Like the previous commentator said, it is nice to have the exposure available to the masses, but the claims made by Prof. Koller are WAY off base!

Chad Colgur

In response to the professor commenting on 23-Feb:

I don't see any claim that "[...] computer-gradable exam or homework assignment can accurately measure mastery in databases or AI [...]" in this article. In fact, if you take a little time to read about the proposed certification process associated with MITx (which, by the way, is the only institution talking about objective assessment rather than a simple statement of completion), then you will see that it is (1) fee-based, and (2) in-person/on-site.

Can you really say that you haven't had a single multiple-choice/right-wrong/yes-no question in any of your course material over the years? Sure, that may only be a small part of your assessment process, but I dare say that it must play a part.

I think it is more productive to think of this format as a start for would-be students. It lowers the barrier to entry. A motivated, self-directed learner like myself receives orientation on a topic that might otherwise be difficult to approach. A student interested enough in a topic to enrol will be able to make a more informed decision about pursuing any given topic further after one of these courses. I believe that is the true value of the format.


I took the machine learning course last year and am taking the PGM course now. I really appreciate the professors' effort to make the courses available online for all to take. It helps with my current job.

Displaying all 4 comments

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account
Article Contents: