The first generation online courses I've taken are still in a very rudimentary form, perhaps because they are quick and dirty ports of real-world classes. They recall some of the first generation mobile apps which were simply wrapped websites. If you were to build a class from scratch for a MOOC, as an experience, you'd make many different choices than if you were trying to just convert a class that was taught in the traditional sense. Just as native apps have come to dominate the mobile phone experience, the next generation of online courses, built from the bottom up to be an online experience, will make today's online courses look primitive. The issues, of course, is that the tools are limited so far, and building such an experience is a high upfront fixed cost.
Companies like Coursera and Udacity need to arm instructors with the tools to build a great experience at a lower investment of time and money. That will come. The benefits come downstream: once constructed, an online course can be spread to a near infinite number of students at a fractional marginal cost. This has always been the chief appeal of online instruction to me, the ability for great instruction to achieve massive leverage and scale.

That was an excerpt of a post I wrote in May on MOOC's. Chris Dixon cites an MIT professor's lament about the current state of online education and adds:

Online courses are to offline courses as movies are to plays. The marginal cost of delivering online courses is minimal. The potential audience is everyone with a smartphones and an internet connection – about 1.5 billion people today and growing quickly. There is no reason we shouldn’t be investing as much to produce online courses as we do to produce Hollywood movies.

The MIT Professor, Woodie Flowers, is candid about his assessment of MIT's OpenCourseWare.

In the United States, our “education” system is choking to death on a failed training system. Each year, 600,000 first-year college students take calculus; 250,000 fail. At $2000/failed-course, that is half-a-billion dollars. That happens to be the approximate cost of the movie Avatar, a movie that took a thousand people four years to make. Many of those involved in the movie were the best in their field. The present worth of losses of $500 million/year, especially at current discount rates, is an enormous number. I believe even a $100 million investment could cut the calculus failure rate in half.
Why not OpenCourseWare?
I argued that the program that became OpenCourseWare should have focused its original $100 million estimated budget on two topics. I suggested microbiology and electromechanical systems as examples. Had we done that, I believe we would have accelerated changing education. We decided, however, to assume that the world could hardly wait to see our huge pile of PDFs, PowerPoint presentations, classroom locations, teaching assistant lists, and other assorted bits of information about our courses. We now have a large database developing digital rot and becoming increasingly irrelevant. It is unlikely OCW will be systematically Facebooked, or Twittered, or HTML5ed, or deFlashed. It is an expensive and unsustainable “free” system.
We have spent about $40 million over 10 years. Powered by MIT’s incredible brand recognition, OCW has made an impact and been celebrated with awards. About seven years after OCW was launched, Salman Khan, our next graduation speaker, started posting a coherent, concise set of tutorials that were inexpensively produced but backed by a pedagogic philosophy. When I last checked Google Trends, the Khan Academy’s search hits exceeded OCW’s by an order of magnitude. Khan designed a product that teachers and students want and need. His modestly-produced presentations are used by millions. Starting with zero brand recognition, he has matched or exceeded OCW’s impact. What might we have done with $40 million, 10 years, and the most powerful technology education brand on the planet?