For those of us with kids and without access to a world-class university this is huge news. I had to study in Spain. Becoming and engineer here was difficult, but the difficulty laid not in the things we had to learn, but in the hoops they made you jump through. With a few exceptions ---there are some great teachers here too-- it was more survival training than a learning experience. Later, when working for HP, I had the opportunity to take Stanford graduate courses. The quality of the teaching was mind-boggling. How lucky these guys are, I kept thinking. Such opportunity!
Of course I jumped at the chance of taking the first AI class, eager to see if it would go towards the (possibly hugely profitable) world-wide higher education framework I was hoping we'd manage to build. I was disappointed by the certificate at the end only mentioning Stanford to make sure that we didn't think this was associated with them in any way. They were clearly not getting it.
But Sebastian Thrun was, and I think Anant Agarwal at MIT is as well. The extremely high teaching standards they have already shown: add a way to give trusted certificates, and this could be the way out for millions of kids throughout the world. And, I assure you, people everywhere _will_ pay for a paper from a trusted source that shows the degree of excellence they or their kids have achieved.
The current Udacity exams take about 10 hours of coding to complete, so it sounds like these are actually going to be significantly less difficult. Still, I think this is a step in the right direction. Of all the online education programs out there my money is definitely on Udacity (or at least the Udacity model) to be the winner, and I think ten years from now people will recognize that this announcement was every bit as big as the X-prize or the recent SpaceX docking.
That's not to say there isn't room for other programs that are run in different ways, not only will there be but many will be insanely successful and profitable, but I think that the Udacity model is going to sort of the default way of doing things that every other program is compared against.
You prefer Udacity to Coursera? I thought the opposite. Udacity has a small number of courses and they are focused on a specific project. Coursera mirrors a university course and tries to broadly cover the subject.
I haven't actually taken an Coursera courses yet. But I generally dislike traditional university courses and I hate Java, so I probably will avoid it until I run out of Udacity courses to take.
A bunch of their courses (including their algorithms courses) are in Java, which makes me skeptical of how good their taste is. It's a strong red flag that they're just licensing content without any understand of the actual content matter and the quality of what they're teaching. I could be wrong, but that's definitely the impression I get based on both the fact that they're using Java and the classes are structured like traditional university classes, since not only are most university classes really mediocre (at best), but it also shows that they aren't taking advantage of the power of the platform/medium. Again, another red flag for poor taste.
Your reply is hilariously naive and insane. Have you even looked at it? It's Stanford professors, who also teach Stanford versions of the same courses.
Andrew Ng, director of Stanford's Artificial Intelligence lab, also the associate professor teaching "CS229: Machine Learning" and "CS221: AI" himself teaches the Coursera Machine Learning class.
Alex Aiken[1] teaches compilers. Tim Roughgarden[2][3] teaches the algorithms class.
And you are going to say they have poor taste and are just licensing content that they don't understand?
Java being one of Google's 3 official languages must be a red flag of their poor understanding and taste too I suppose :)
NLP class allows Java or Python - this is because they need to provide framework to perform auxiliary task, and students are required to fill only needed parts - specific algorithms, etc.
Coursera is pretty language neutral. I did Algo I and used Python. Other students used C, C++, Java. And when I did Machine Learning the recommended language was Octave/Matlab (but you could use Python if you know what you're doing).
"The current Udacity exams take about 10 hours of coding to complete..."
From the article: " For the first round of exams, programming will not be included."
I don't think you can demonstrate programming competence with a 90 minute multiple choice test, so I'm interested to see how they end up tackling that challenge.
I think the idea would be for the Pearson exam to act as a compliment to verify your identity more so than demonstrate coding competence.
While the Udacity exam is the actual in depth proficiency challenge that employers will care about, the Pearson name gives some sort of assurance that this person did indeed pass a difficult coding related exam in a physical setting where we can verify their identity, reducing the likelihood that someone cheated or had someone else complete the Udacity course for them. The actual test questions might ask something like "How would you approach problem x, describe what language you would use and how you might structure a program," or for multiple choice, "Which of these is written using proper Python syntax" etc. If you have to learn enough to pass the Pearson test in the end, you'll still have to put in a large chunk of time, making cheating downright impractical.
End result, this partly addresses one of the fundamental issues with online learning that Sebastin Thrun has talked about before, providing a form of physical identity verification and association with an online student username. Of course, someone could always pull a bait-and-switch in the physical testing center, but that's a whole other ball game.
Learning for these kind of tests was always the most pointless tasks in school. It is just the same extreme-memorization, braindumping, completely forgetting about it pattern. But of cause this structure makes any kind of course super scalable.
"...students wishing to pursue our official credential and be part of our job placement program should also take an additional final exam in a Pearson testing center."
It sounds like the Pearson exams are in addition to the online, coding test.
Interesting, it's a little ambiguous as to whether they mean a second final exam or whether they mean in addition to the regular course. Assuming they mean a second final exam though this would be good as a clever checksum just to make sure you are who you say you are and can answer basic questions.
This is fantastic and a required piece of the online learning revolution. You have to prove that it's you and take an exam to get the credit and currently this can only be done by physically going to test centers. I predict that in the very near future (~1-2 years) taking such tests to get credit for online courses will be as common place as taking the GRE or GMAT.
I'd rather see web-cam based proctoring or a similar solution that would allow for some accreditation while letting udacity still control costs and provide a cheaper overall examination process to its end users.
The endpoint is a model that will satisfy the employers - whether that's webcam or live proctoring or any combination thereof is irrelevant.
One variant of live proctoring could be an arrangement in which universities give students a number of online options to fulfill a particular requirement - cs, math, physics, etc - and administer their own exams. So the university would simply use it's brand to issue the stamp of quality. In a sense - this is cannibalism for the university. But I think there are public schools that are struggling to serve their students in a financially feasible way. This could be a good way to streamline...
I think it was vWorker, but you're right. They have a time tracking tool which randomly takes screenshots, and webcam shots. Now I don't know about screen shots, but for Udacity the final exam tracker should be allowed to randomly take webcam shots.
Maybe there will be an added inconvenience of opening the software, maybe there'll be people who might game the system. But for a large part of the population I think it should work
I agree, by outsourcing to test centers they are side stepping a giant problem. I hope it's a short term solution. I loath the thought of testing in a clinical setting for a stressful exam, there has to be a better way.
I'm reminded of the old microsoft certification hoops I was required to jump through for a previous job (I think I ended up with about 30 or so of the things and a load of letters to put after my name that I now cringe to think about, mainly because of how absurdly pleased with them I was at the time. HR departments and management were impressed by it, the people who worked at the coal face, not so much.
Be interesting to see how it turns out.
EDIT - and of course for things slightly more academic, I'd expect there to be not quite the same set of issues as there were with vendor qualifications (before someone points it out). Yes, I too would prefer to hire people who maintain an interest in learning over someone that didn't (I'm signed up for four of these things as soon as my MSc is out the way), so not intended as a derogatory comment.
I hate Pearson products with a passion (save for a few imprints and assorted acquisitions). This is a great step though for both companies, so hopefully it works out for both. Pearson's online components has included some of the worst software I've ever had to deal with– maybe thanks to this, future students will never have to deal with it again. Udacity on the other hand has been great, so it's all very exciting anyway.
Pearson are just a big acquisition/sales company so have some good/some bad products. Their model of business is not where I would personally hedge my bets if I wanted my product to be successful. Udacity probably have no idea how difficult it will be to work with Pearson yet.
If the partnership gets it right then Pearson will look to acquire Udacity, given Pearson's track record of how little they innovate after acquisition this may be damaging to Udacity's long term goals however if the stakeholders of Udacity are looking for a quick out of the market then it might be ideal for them.
The Pearson VUE exam driver is pretty solid (Java). But I agree, all the ColdFustion stuff Pearson does is very stinky. VUE was an acquisition I believe.
Inversion of the classroom is an idea whose time is come. In the university setting it's easier to do than in a junior high school setting. People can all learn at their own pace. And then schools will primarily do testing and accreditation.
The question is, should the lesson plans (for the overall year) be open source, or a marketplace of proprietary competition, or both?
Awesome. Aping the accreditation, credit and degree system of brick and mortar academe doesn't make sense for the latest generation of educators - udacity, Coursera, Khan Academy, et al. Proctored exams to verify that the student knows the material are the way to go.
For final exams that cover an entire course, a test independently designed by a third party, using input from employers, entreprenuers and academics, might be a great supplement. Where you took the course would be irrelevant, so long as you could pass the test. It would allow for easy comparison of students in different programs.
Testing is a vital step in providing a credential at least as valuable as a college degree.
Of course I jumped at the chance of taking the first AI class, eager to see if it would go towards the (possibly hugely profitable) world-wide higher education framework I was hoping we'd manage to build. I was disappointed by the certificate at the end only mentioning Stanford to make sure that we didn't think this was associated with them in any way. They were clearly not getting it.
But Sebastian Thrun was, and I think Anant Agarwal at MIT is as well. The extremely high teaching standards they have already shown: add a way to give trusted certificates, and this could be the way out for millions of kids throughout the world. And, I assure you, people everywhere _will_ pay for a paper from a trusted source that shows the degree of excellence they or their kids have achieved.