Part of the backstory here is that ai-class.com is by Udacity (Sebastian Thrun), while ml-class.org and pgm-class.org are by Coursera (Andrew Ng and Daphne Koller). Formerly colleagues from the same department, now competitors with very similar education startups, all the way down to the naming conventions. Lot of fur flying about who copied who.
Coursera has been launching a ton of classes[1]. Probably Sebastian feels that to beat Andrew and Daphne, he has to go full time.
" Formerly colleagues from the same department, now competitors with very similar education startups"
(Without taking sides, and somewhat tangentially) if true, we consumers are in for a good time. Competition,especially with such formidable people leading rival companies, is awesome.
So, does this explain the unexpected delay of the Spring ML-Class? I haven't heard any news until a few days ago when I got an email saying there were delays in starting the new semester.
How did this happen - did he want to push online education in different directions? Did he want to pursue more overtly commercial interests than his counterparts, which seem to be approaching it from a nonprofit sense? Was there a personal falling out between Andrew Ng and Sebastian Thrun?
The non-profit focused alternative seems to currently have both a better platform and more participation. Can a lesser amount of potential revenue allow this to scale up? Or will professor's be drawn by the larger salaries commercial teaching organizations could command, coupled with more extensive and rapid feature deployment?
I know of at least one Stanford student who feels swindled by this transition to online classes. 1. For some stanford costs a lot of money while online it's free and 2. The time taken developing this class may take away from the in person experience, further reducing the value of attending classes at Stanford.
Any idea if these feelings are wide spread? I only have the one sample.
I find this attitude really strange. The value in going to Stanford isn't that you get to take the physical classes, it's that you can walk up to Prof. Ng's office (and/or similar caliber profs) and have a chat, and more importantly target yourself to be doing some research under them before you graduate.
Having merely taken a class with someone big in the field (especially a large class) is essentially an interesting point of trivia to bring up at dinner sometime. If anything these online courses are about demonstrating that the scarcity of chairs in a great professors classroom is rather artificial. Having done research under (even if it's pretty trivial stuff you personally do) someone great will really have an impact on your career.
If you're an undergrad and your not spending a good amount of your time talking one on one with researchers you have access to you're not taking advantage of a huge benefit of attending a university.
The online students aren't getting the most valuable part of the entire thing - a diploma with the name "Stanford" on it. Yes computer programming is a young profession that isn't as hidebound as many others, but it is maturing and as is true for most other professions, soon proxies will dominate hiring.
Even more interesting to me anyway is the fact that this is potentially a sign of a shift in the industry from importance of the traditional higher education credential to importance of a worthy alternative credential not owned by traditional university systems and structures -- which is extremely exciting.
I still agree that the networking components and group-learning (or face-to-face interaction) are missed (and are extremely valuable), and so it will be very interesting to see how this new paradigm deals with that issue.
I'm currently enrolled in Daphne Koller's Probabilistic Graphical Models, which will be offered later this quarter as online class. The sentiment among my friends is largely positive. First of all, personally I embrace the entire philosophy of moving toward better educational models that leverage technological progress, I am excited to be caught up in the transition, and I am more proud to be a student here. But more importantly, the instructors put more effort into the courses now that they are online, so we benefit directly from better prepared exercises, more of them, better structure of the class, etc. Naturally, there is a difference between how much preparation goes into teaching 50 people in a room compared to teaching 50,000 while spearheading educational revolution.
As another sample, I'll say I'm far from feeling swindled. I was a bit disappointed that the publicly available ML class and its Stanford equivalent only differed in one optional discussion a week - but I certainly don't feel the time taken developing the class took away from the in person experience, and I do think Andrew Ng and the staff were more available for Stanford students than they were for those taking the class online (which I view as the justification for the Stanford tuition costs).
The cryptography class I'm taking this quarter is more to my liking in that aspect, as in addition to the video lectures (which I assume will make up the bulk of the public offering), there is another 2 hours of lecture and a discussion that cover additional material. In this kind of a set up, the added benefit for the Stanford students from an educational standpoint certainly seems more tangible.
you mean Coursera is founded by Andrew Ng and Daphne Koller? well that means Stanford picked Coursera over Udacity, all their other online courses due to start are on Coursera. Andrew Ng is also the Director of their AI lab, pretty unusual for someone who isn't a full professor yet.
Neither of them seem to have a collaboration with iTunes U. That is going to be where a big chunk of page views come from, and I guess Apple even manages the streaming infrastructure for it.
Took the AI class and it was just amazing. I never had a professor more passionate (yet still rational :) about a subject.
Even though the course made the math and the background sound simple, it wasn't. There is a probably thin line between breaking-down things into a set of well-partitioned and easy to understand statements and oversimplifying really complex systems.
Also, the applications (edge detectors, robotic cars, particle filter based localizations, ...) kept me very motivated throughout the course.
having taken all three classes (AI, DB, ML), I have to point out that in my opinion (and that of other people I spoke with) ai-class was _by far_ the worsely organized, passion of the teachers notwithstanding.
But I am sure Know Labs learned a lot from it and they can only improve.
The website itself sucked, the quizzes weren't very usable, but the videos themselves were wonderful — in my opinion the best. And I wouldn't be surprised if that was the part that's hardest to improve on, and also the most important one.
Totally agreed. Prof. Thrun and Prof. Norvig made math an interesting part of the course. At the end of the course I found myself more motivated to learn mathematics than ever before.
Yes, and on a related note: Today morning I skimmed the preface to Computational Science and Engineering by Gilbert Strang, and he also states, that (at least some) math should be thought alongside applications and algorithms, which, since these used to be intense topics, have been traditionally separated by academic curricula.
AI class was a perfect example of that interleaved approach, I think.
It looks like Khan Academy for college-level, semester long classes. He should target the University of Phoenix's crappy online program. Their parent company, the Apollo Group, has a market cap of $7B. Thrun could easily take a huge bite out of that.
While I agree that a new education startup could go after something like the University of Phoenix, I'm not sure that Thrun and his crew are going for that.
To me, the University of Phoenix represents a credential factory. People see that all the jobs they want require a college degree, and so they go the easiest route to get that. I don't believe that University of Phoenix (or most of the other big for-profit names) have a strong interest in educating people, nor do I think that their customers are really interested in getting an education - instead they're interested in getting a piece of paper that says they're qualified to do the job they want.
Thrun seems to have far more noble, though possibly less profitable, goals in mind.
Not with one class and no accreditation, he won't. UOP is the largest private educator in the US with more than 400,000 paid students. Setting aside the fact that a single online class is nothing close to a "University", he's got a long way to go before he can make a dent in the #1 company in the industry.
I have built something to keep track of all these courses http://www.class-central.com. Currently tracking just Coursera(Stanford's online learning initiative) courses, but would be adding UDACITY and MITx courses(when they are announced) soon.
That looks very useful, thank you! I've signed up for a bunch of these courses, but no emails have come in yet and even the websites themselves aren't very connected, let alone the ones on different frameworks.
Great work collecting it. I would do some user-tests on the landing page though. It needs a tiny bit more content and a clearer 'best path' call-to-action.
I think Stanford is more for keeping things closer to the status quo. Students in seats is how they make their money, after all. Not to say that they won't make courses online. I think they would be more than happy to charge $5,000 like they do through SCPD. However, if this develops the way I think it does, the economies of scale have to take over, right, especially after reading about Professor Norvig's discussion last month (http://remotelearningproject.com/interviews/peter-norvig/) about potential business models that try to keep it mostly free.
I'm very excited.
PS: I've noted that the PGM course website says it'll start in Feburary.
I hope this is just a first step toward big future ideas.
Many highly sought professors already have great brands, and I'm surprised that they are so highly underpaid for what they do. Many of them could be getting paid a lot more in scale. Further, they could also be providing value-add services to directly validate some of the best on-line students and grant certifications of expertise. Think of the mozilla badge model [1]. There are hundreds of ways to spin out revenue and scale that model.
The average salary of a professor at Stanford is $148,000. I think it's unfortunate that he had to step away from the possibility of influencing those who do make it to Stanford in favor of catering to the world. It's great that he does so, but there's also plenty of reasons why he would be affecting many different minds at Stanford.
For those interested, looks like they're offering two classes (CS101: Building a Search Engine and CS373: Programming a Robotic Car) starting in February and hiring actively, as well (http://www.udacity.com/jobs).
Prof. Thrun is an amazing teacher, but I think the 160,000 student sign-ups were due to the Stanford affiliation. Giving up that affiliation will cut enrollement sharply. Any likely business model will cause another sharp decline in enrollment. Thrun has all the right ideas, but on his own it will be tough to execute.
Education is ripe for disruption. Thrun+Ng+Norvig+Stanford as a cohesive team could have made a history altering change in education. It's unfortunate that they aren't a team.
It all comes down to certifying student proficiency (course credit).
Udacity and Coursera are not really competitors in the area of course content. They are competitors in the area of student certification. In the reputation behind the process. Meaning rigorous final exams, independently administered, suitable for inclusion on curricula vitae, etc.
By offering courses for free, Udacity and Coursera compete directly with Stanford. But Stanford can compete with them just as well, by allowing students to enroll in the free classes, mentoring them, and then offering their own certification exams -- for credit.
My guess is they won't do that. They'll just find someone else to teach the course. But I bet in the future they intend to keep a very close eye on the department.
Since doing the online AI course,
I have been hoping Thrun would teach Robotics.
I have signed up - woohoo.
It looks awesome - I am stoked - Sebastian Thrun is an amazing teacher, he really makes me think hard and gives me the scaffolding to investigate further on my own.
What I find problematic about all of these new online education startups (Udacity, Khan Academy, etc) is that they tackle the problem by simply providing online equivalents to traditional didactic learning methods like lectures and textbooks.
A professor standing in front of a group of students lecturing is definitely easy for the professor and cost-efficient to scale up to larger class sizes, but that's just not how a lot of students learn. I'd wager that most HNers learned programming through actually writing code, even those who learned CS through a formal program. In the humanities, I'd argue that the most effective way to learn is through small discussion groups, not a distinguished professor explaining literature or philosophical works to you. Just throwing that up on the internet is an easy way to expand your audience, but providing higher-quality educational materials doesn't do anything to improve the quality of how we educate.
The internet has a lot of potential to improve the quality of education, and there are tons of awesome startups working on it (companies like Codecademy and Coursekit come to mind), but I personally hope the future of online education doesn't look too much like Udacity.
I agree, but it does provide value beyond a simple real-life lecture, most notably the fact that you can go at your own pace and rewind/replay at any point. This is huge. In addition, the lectures are much shorter for the most part, which also makes a significant difference.
Concerning the importance of discussion groups, face-to-face interaction, and networking opportunities -- those things are necessary in certain cases, but not all. The courses currently being offered through these sites are most often the ones that don't necessary benefit hugely from these real-life components.
I don't think they're discouraging the self-discovery component of education. Even Khan Academy's holistic model seems to treat his videos as just the bare-bones - the site encourages lots of practicing on the website and projects. The ML class encouraged coding projects, and Khan Academy's "schools" in CA involved inverting classrooms such that projects and self-discovery was performed in the class while the bite-size lectures were studied outside of class.
I wonder if his departure has anything to do with the delay of the other courses? Maybe he took some students/staff/resources with him that were key to the infrastructure?
The other courses (i.e ML and DB) seemed to have a different arrangement. The slew of new classes seemed to be following the same arrangement as ML/DB rather than AI.
Though as far as I know there was this idea of using Aiqus as the discussion forum platform for all the courses, rather than each course having its own separate forum. Aiqus was used by AI class during last session.
Although I am not aware who was behind the development and operation of Aiqus.
> there was this idea of using Aiqus as the discussion forum platform for all the courses
Not sure if you're referring to the current bunch of courses or the first three (AI/ML/DB). I'm assuming it's the latter. Aiqus was only used for the AI class. IIRC it was set up independently but someone before the course began and ended up being the pseudo-official Q&A forum (linked to from the main site).
ok, I was not aware that Aiqus was independent of the AI class. Though one of the developers of Aiqus posted in an unofficial FB group of AI class that all the rest of the courses will also use Aiqus during the next session.
Coursera has been launching a ton of classes[1]. Probably Sebastian feels that to beat Andrew and Daphne, he has to go full time.
[1] http://www.cs101-class.org/hub.php