It took most professors completely by surprise, and it sounds like there may not be bottom-up buy-in (was completely a management decision). In particular, at at least one university, profs are now being told that they may be required to teach a course via Coursera, or at least strongly requested to. And, the universities don't want to budget this as actual teaching, so it's just extra work on top of the normal class load: since Coursera courses aren't credit-hours, they don't give teaching credit. It's supposed to count under "service" or "outreach", I guess, the way serving on committees or reviewing papers or doing a CNN interview does. (This part may vary by university.)
Not sure that's a good recipe for high-quality courses via this method. The advantage of the first few courses is that it was a bottom-up decision by professors who wanted to do it, and devoted significant time to do it right, rather than having it assigned to them by management.
If it is true however, there is still udacity, which is much more conservative in it's expansion.
We'll see, though. One thing universities are trying to balance is how to get professors to do more "entrepreneurial" kinds of things while staying. As professoring gets to be a little more like a freelance job (bring in your own research funding, teach online courses via a for-profit company, build up a personal "brand"), one question is what keeps the more successful professors at the university, instead of just going independent, the way Sebastian Thrun did. There's a real possibility that some of the launch professors listed in this announcement volunteered precisely because they're contemplating jumping ship to their own online-course startup, and are using this as a route to build name recognition and audience.
Example: many course videos are of a "live" class and are repeated for every semester regardless of the subject. Next to no re-use of course video is seen from semester to semester. The experience is very much one of time/place-shifting a class experience.
Also, the focus is almost NEVER on interactivity. As a distance student, you're VERY much dependent on your professors willingness to read and answer emails or post in forums.
I was very surprised that MITx had an IRC server set up for the EE class that were offering. I nearly fell out of my chair when I saw they had hundreds of people in the channel.
So, from my point of view, the biggest shift we are seeing is a willingness of the brick-and-mortar universities to consider working with a partner that is going to market their professors and university brand to a very wide audience and yet enough of an arms-length away that there is little risk to the "brand" if things go sour. The fact that there are NO for-credit courses is quite conspicuous and you should expect the universities to attempt to maintain an implicit separation between MOOCs (massive open online courses) and their traditional "residential" offerings.
That said, the writing is on the wall: the traditional model doesn't scale and financial pressure is going to push ALL schools towards this model in some fashion or they will simply become irrelevant. As I drive up and down I-95 in the Northeast, one of the things I've noticed is that there are a massive number of traditional schools advertising on billboards. This NEVER used to happen and most universities (IMHO) considered it a sign of desperation to advertise for students as opposed to recruiting.
What we are witnessing is a race developing. Think back to the early days of Yahoo/Altavista/Google. There are going to be some big winners and some also-rans in this fight. I'm betting on it.
Full Disclosure: I earned my bachelor's degree from Harvard via their Extension School and did the majority of my coursework online using their video delivery platform. I graduated in 2009 at 39 after a LONG absence from school to chase my fortune on the Internet. Best decision I ever made.
I don't disagree there are all kinds of problems. But how does this initiative address any of the financial pressures? It does nothing to reduce the cost structure of universities; in fact it increases the cost structure by adding another thing their professors and administrators must do, not instead of but in addition to everything they currently do. And yet, it does not provide any new revenue streams to pay for that. At least, it doesn't unless there is a payment from Coursera to universities as part of this deal that hasn't been announced yet. My guess is that they're hoping to use it as an advertising loss-leader to attract students and prestige. But that would mean doubling down even more on the traditional high-tuition model, because that's what's going to ultimately subsidize the free courses.
Now if they charged for the online courses, I could see that making a difference in the financial picture. Maybe that's the longer-term plan, for-pay online courses with Coursera and the university splitting the proceeds. That would be closer to the Harvard Extension School model you mention.
The revenue model is going to be something that is roughly similar to the ones we have with any number of "freemium" type services.
My bet is that there is going to be a certification track that complements the online course delivery. I've seen this with industry certifications (JNCIE, CCIE are ones I'm familiar with) and that the cost to get "certified" will be fairly substantial. It'll also require physical presence and some practical demonstration of competence. For math, I'd expect a traditional sit-down exam. For CS, probably some combination of programming exercise and exam. I don't know how this will work for the humanities.
And one other thing considering the cost argument: paying for the production of a video lecture and the hosting of the content for hundreds of thousands is going to be a fraction of the cost of paying even a grad student to give a lecture on-campus.
Mark my words: in a few years, you'll be able to certify that you passed each course and you'll get some form of certification for passing a sequence of courses. It might not be called "BS in CS from MIT" but it'll be looked at the same way.
But a lot of people are just undecided, waiting to see how it'll pan out. Probably not a lot of outright enthusiastic professors, except from the relatively small number who are actively looking to either jump to one of these startups themselves (like Sebastian Thrun), or at least co-found one (like Daphne Koller).
The Chronicle of Higher education's occasional guest op-eds are probably a decent representation of the range of opinion. Here are a cautiously positive and a somewhat negative take, respectively:
I think a lot of the academic scepticism does come from the for-profit - not to say for-profit education can't work, just that it is somewhat new territory for many institutions.
I signed on for Daphne Koller's Probabilistic Graphical Models and Geoff Hinton's Neural Networks courses. PGM is supposed to be really tough. I am planning to take a couple of months off from work.
This new world, where you can learn for free from the experts, is a dream come true.
I am thinking of blogging my progress (or lack of it, as the case maybe!)
I'm not sure that's what they do though. Learning "from the experts" means they are there to help you, understand where you're coming from, and correct you along the way. Just watching a video of someone with recorded slides is something we could have done 20 years ago.
I was excited about the PGM class, but my interest dropped after the first few videos (100% comic sans slides, poor video quality, poor audio quality).
Just because you are an expert in some field has zero bearing on your ability to teach or convey information. I'm not sure a "let's imitate college lecture formats" site founded by smart people understands that. They fall into the trap of "smart people can do EVERYTHING!" There are a lot of very smart people who have negative presentation skills. There are a lot of very smart people who have excellent presentation skills too.
Solution: have the smart people write the script, but have people excited about presenting show it to the world.
Ultimately, I believe, there will have to be different (types of) lectures for different people, and different speeds.
The UI, however, could be improved.
I do like the programming interface that these sites use, to work out problems and get instant feedback.
Would you consider yourself an auditory learner? If that's your preferred learning style, I guess I see the value in these recorded lectures.
Anyway, I wanted to see what one was like. Since it's free of charge, no big deal if I don't care for it. But I'm curious.
In a Udacity course I took (the driverless car one with Thrun), they did have a (community-generated) html version of the lectures, but you would still have to go back to the videos to get credit for having watched them and answered the questions -- and, of course, it was always a work in progress.
The advantage over textbooks would be in a) hyperlinking, and b) a known group of people doing it at the same time as you so you can discuss the material and get answers from the professor on the most common questions.
Having said that, I prefer learning from notes. Not text books, because they are usually way too verbose, but during my studies, I would usually skip lessons (and save 1.5h on commute), and study using my classmates' notes.
I personally think there's room for many services like coursera and udacity to serve different levels of academic experience and expectations.
The lessons are interesting, and I really like their concept of quizzes in the middle of the lessons. They just show the question on the video, he explains it, and they stop the video and you can choose which one is right, or type in the number.
The homework is also well done and enhances the knowledge from the course. I had a lot of fun doing the lectures each and every week. You probably want to put aside a set day and time when you will do it, though, or else it is easy to do other things and "do the lectures later".
If they continue like this, there is soon no reason to go to a university anymore. :) (which is a catch-22 since the lecturers are payed by universities...)
For example, whenever I do the washing up I put my tablet on the windowsill and watch a video. I've also been watching them instead of browsing the web as a more structured break when studying. The fixed length of a video makes it much easier to be disciplined with the length of my breaks. The subject matter is different enough and presented in such a digestible way that it really is a good break.
It's just alpha nerding, same as any other venue. And if it's really that easy, then as you say, why not impress us with how you've gone beyond it. Anyone who can type can say it's easy. Show us something.
I was just being honest. The exercises were easy for me. I'm sure they were not for many people. I don't care, because I don't compare myself to others in this way. In comparison to other people around me, I know I'm very smart/capable for some things, and suck at others. I'm not saying this to feel better or so that you would respect me more. I'm saying this because it's fact. The way I see it (without having to compare myself to other people), I'm capable enough to reach many goals in my life, but I would like to improve myself in many areas still.
But anyhow, you've completely missed my point.
In addition, vectorization wasn't even necessary to solve the problems. Yes, that's how I too made the exercises more interesting, but that could be done with almost any kind of programming exercise in Matlab.
If it's concentrating more on the math than on applying the stuff to real-world problems that bothered you, then you wouldn't want to take the real full-blown Stanford class. That class is largely doing rather difficult math proofs and the like.
I'm not sure I understand the argument that you could have solved the problems in a lazy fashion, however. Such lazy solutions aren't good for the real-world, as they don't perform well-enough. Part of the insight taught in the class is that you have to look for these sorts of performance optimizations in order for the solution to be feasible to use in the real world.
If the problem is that other students could pass the class without having put in the same effort as you and I did, who cares about that!
I studied math. I love math proofs. It's just not what I would expect if I would sign up for a ML class. Note, however, that I'm not complaining about the contents of the ML class, which I think were awesome, a really nice broad overview of the ML techniques. I really think I learned a lot. I'm just complaining about the homeworks, nothing else. They were not ML homeworks, they were CS homeworks (particularly, Matlab class homeworks). I don't know the contents of the real Stanford class, but believing your description, the same kind complaint applies.
You're wrong when you say that "lazy solutions" don't perform well in the real world. In the real world, robots aren't programmed in Matlab (I hope). In compiled languages, loops are just as fast as vectorized code (which is implemented using loops). Vectorized code having better performance than imperative looping code is just an artifact of the fact that Matlab (and e.g. R) are interpreted languages. (Sure, there might be some additional optimizations possible on vectorized code, like SIMD instructions, cache locality, special algorithms for sparse vectors/matrices, but this is in most situations premature optimization. Being able to write correct code and read it later beats 2% gain in performance!)
If you really want to understand ML, you should do the real class. All the material is online, so you could do it at your own pace.
The online class, however, was a watered-down version of this, and designed to give you a taste of the real class without being brutally difficult. Perhaps it will inspire some people to do the work of the full-blown class.
On the other hand, sometimes you just want to know how to apply some technology to real-world problems without having to understand all the gory details. I think the online AI class is more along these lines. They had a final project, for instance, that was structured as a contest to see whose solution worked the best. I think they used canned ML packages instead of coding them on their own, but then wrote their own code to apply the canned ML software to some sort of open problem. Perhaps you would enjoy that class more.
As to getting only a 2% performance gain from vectorizing your code, I'm skeptical of this claim. I know that highly respected people (e.g., Guy Steele) have claimed that people still use Fortran, despite its drawbacks, in part because of it superior performance over C in being able to vectorize code and getting substantial performance gains. Furthermore, a vectorized algorithm is more easily adaptable to a Hadoop cluster, or a GPU, and both of these can lead to huge performance gains.
In higher-level languages, such as Java, vectorized solutions can also lead to much higher performance, as you can use a native linear algebra library, which will, no doubt, perform better than Java loops.
Furthermore, if you can turn an algorithm into linear algebra, you will have a much more maintainable solution, as you can turn a program that is several pages long into one that is several lines long. That is surely going to be easier to maintain!
I've programm(ed|ing) a spaced repitition native app that uses ideas from online sequential learning to challenge me (hah online learning for online learning). I can look at performance graphs per topic area over time. For now it just scores roughly based on text similarity but I've built a parser I can use for the more mathematical ones. By the end of a couple of courses it should be a very interesting project. It currently additionally allows for searching and note taking but I noticed when I go over my notes I tend to go back and forth on certain concepts so I will implement a history concept and automatic topic map building (using both cookie trails and measures of similarity). An unintentional side effect due to Markdown support and html generation is that it generates something like a wiki but far easier.
My approach is inspired by http://en.wikipedia.org/wiki/Roger_Craig_(Jeopardy!_contesta...
I'm not enough of a hacker yet to actually do that myself
After aimlessly reading HN and blogs for so long, having a solid, structured course feels so... good. I need to mentally prepare myself for this very real thing called "learning" once again!
Personally, the old saying
Not only is a book the better teacher, it also has more character.
holds mostly true for me. Most classes, that I remember to be superior to the experience of a good book, were either because there simply wasn't a good book, or the teacher was extremely charismatic, or both .
However, I really like this movement, see its appeal and learnt a thing or two from these new , and old , online courses.
As others have pointed out here, watching a short video has very little transaction cost. That is true, however, for me the transaction cost with a book is often lower. With a book I can easily navigate to the interesting part. With the online videos, I never know if I can skip over the intro stuff.
For instance, in Ng's ML class  a lot of the material, especially the majority in the first six chapters was familiar to me. However, had I not watched the first chapter I still wouldn't know the definition of a Hypothesis in Machine Learning. This problem exists with books, too. But I think skimming is easier with books.
 Underactuted Robotics http://ocw.mit.edu/courses/electrical-engineering-and-comput...
I easy to buy a book on an interesting subject (I do it all the time), but it's much harder to actually read it, especially if it's on a challenging topic.
For me, the best solution would be if the courses had the material available in text form as well. Then you could choose how to learn (video or text - I would choose text).
I'm trying to get used to the model by taking courses that will mostly be for review (but I expect that they will get into some material that is new for me, or forgotten through lack of practice for me) such as
Statistics One by Princeton University starting 3 September 2012, taught by Andrew Conway
Calculus: Single Variable by University of Pennsylvania 27 August 2012, taught by Robert Ghrist
Introduction to Logic by Stanford University 24 September 2012, taught by Michael Genesereth
I'm currently in the Udacity introductory courses on statistics and physics, again for personal review and also to check them out for my children. I'm intensely curious about how this course delivery model will develop, so I'm jumping in to use the Massive Open Online Courses to study subjects I like to self-study anyhow.
All I can say is that for this course I would even be willing to invest myself even more (both in time and money) at the level of requirement of a graduate course. I really would be interested in paying the standart fee and follow the same course than the stanford students follow and have a kind of official credit in reward.
The problem is that there is basically no "service" at all. The first iteration was delayed by several weeks (without any prior notice or status update) and there were several problems with the certificates (Ive gotten less programming credits than I should have and others had similar problems) but got no chance to get this corrected.
I guess thats the problem with free courses (and thats the reason I would like to pay some fee for it): You can just record the lessons once and afterwards repeat the course how often you want with very little updates - and very little help/service for students.
That's the real challenge to all the middle-of-the-road schools out there. Now they can be directly compared to the best.
What's the business model though? Because that can only go so long without the teachers/experts earning money.. and if they earn money, someone else must be also earning the money to pay them.
I've tried the sociology one but really didn't dig the first texts we had to read.
I'm waiting for gamification, python, object in french, crypto II, greek mythology...
I'm also taking Algorithms in Udacity, but I don't really like Udacity too much.
Both were great experiences, and I learnt a lot. However, each course was many hours of work per week for me. There are easily 10 courses I would like to take from the current offering, but for me (have a family, working full time), the biggest constraint is time. Nice though to have the choice once you decide to take one.
I have written in more detail about my experience with Coursera: http://henrikwarne.com/2012/05/08/coursera-algorithms-course... and http://henrikwarne.com/2011/12/18/introduction-to-databases-...
I took a Coursera course last month, and despite a couple of weirdnesses, it was a really enjoyable experience. I don't really get Udacity, but I'm definitely going to take a couple more of these courses.
It also appears as though they're starting to figure out how to mark essays, etc without involving huge numbers of lecturers. It's an exciting time for eduction.
EDIT: Oh, nevermind, I've found it. It seems that student grade each other's essays - one student grades three other students.
Look here, under Peer Evaluation:
The workload looks reasonably significant, which I'm happy about. I'm a little skeptical of peer grading of essays, but it's something they've got to figure out for the future.
Every class I've done so far has had a tiny minority of forum whining about grades. I'm expecting it to be notably higher in this class, but we'll see how it goes...
Coursera seems like bunch of interesting courses put together, rather than structured degree thing.
I think this is because universities see coursera as way increase its reputation with the public rather an alternative to university lectures.
Udacity seems more approachable and familiar. Oddly enough, even with 2000+ episodes, so does Khan Academy.
There is one that teaches Django?
It is one of those topics that there is not too much helpful material online - so I am looking forward to it.