Is Sebastian Thrun's Udacity the future of higher education? | Hacker News Reader

Is Sebastian Thrun's Udacity the future of higher education?

edition.cnn.com

41 pointssubbu37 comments

No.

The class Sebastian Thrun taught on AI, was a class devoid of all the qualities that made CS 221 one of the (formerly) greatest classes in not only the CS department, but all of Stanford. His AI class was dumbed down and slapped with a Stanford logo to make people think that they were performing at the same academic level as top notch university students when in fact he had simply lowered the level of achievement so that anyone could take the class. AI is hard. Machine learning is hard. Computer science is hard. Not everyone can do it, and no online course will change that fact.

Why can I say this? These are class reviews of CS 221 from Courserank before and after Sebastian Thrun made it his flagship for online teaching.

============================================================

BEFORE:

============================================================

Note: When taught by Andrew Ng the class received no ratings less than 4/5 stars

"4/5 Stars

Autumn 2006-2007

Andrew Ng

A+

0 of 0 people found this review helpful

Covers a broad spectrum of topics in AI. If you are interested in AI, but you aren't sure what area in AI you might want to take classes in or you don't know much about AI, this is a good class to take. After CS 221, you can go on to CS 229 (machine learning), CS 223B (computer vision), CS 224N (natural language processing), etc. This class is a lot of work, and most of it is valuable although not all of the programming assignments were that well designed when I took it."

------------------------------------------------------------

"5/5 Stars

Winter 2008-2009

Andrew Ng

A-

0 of 0 people found this review helpful

Good class. Andrew isn't the most exciting lecturer, but you'll learn a lot of different AI techniques, and the programming assignments are interesting. The problem sets and midterm are heavily algebra/proof-based, so be prepared. Work through the section problems and you should be fine with that.

Since all psets, assignments, and the final project can all be done with a team, make sure you have at least 2 other people you know you can work with, or else you'll get dragged down."

------------------------------------------------------------

"5/5 Stars

Winter 2008-2009

Andrew Ng

0 of 0 people found this review helpful

I really enjoyed this class. Very interesting topics, long and involved problem sets, and not-so-difficult programming assignments. Except, of course, for the final project. GET A GROUP ... I had to drop down to CR/NC because my partners dropped the class, and so I spent most of dead and finals weeks working on this stupid robot dog.

That being said, I loved the class."

------------------------------------------------------------

"5/5 Stars

Autumn 2009-2010

Andrew Ng

A-

0 of 0 people found this review helpful

Terrific class with great lecture material and interesting videos. I think the Audi parallel parking itself by driving backwards at 40mph, braking, and sliding into the parking spot was what kept me going.

The problem sets make sure you really understand the material, and the programming assignments are a great way to learn Matlab. The project is HARD and time comsuming, so make sure you have time in your schedule near the end of the quarter!"

============================================================

AFTER:

============================================================

Note: Under Sebastian Thrun the class has received no rating higher than 3/5 stars, even more telling, look at the comments for Autumn 2011/12 when the online system that Udacity is based off of was rolled out for Stanford students.

"3/5 Stars

Winter 2010-2011

Sebastian B Thrun

A-

5 of 5 people found this review helpful

With Thrun class had a very different feel than it would have Ng. It skimped on the math/theory and focused on intuition and practice. I liked it less, but for people who are less interested in the math, it was an improvement."

------------------------------------------------------------

"3/5 Stars

Winter 2010-2011

Sebastian B Thrun

A-

5 of 5 people found this review helpful

If you can take this class with Andrew Ng I would recommend it. The version I took was pretty poorly taught. The lectures lacked both detail and clear explanations of the concepts. I feel like I came out of this class without having learned much of anything."

------------------------------------------------------------

"1/5 Stars

Autumn 2011-2012

Sebastian B Thrun

7 of 7 people found this review helpful

Hands down the worst class I have taken at Stanford. Terribly taught, unresponsive and incompetent TAs, mixed up deadlines. Avoid at all costs."

------------------------------------------------------------

"User had not rated this course at the time of reviewing

5 of 5 people found this review helpful

god awful. learned absolutely nothing."

------------------------------------------------------------

"User had not rated this course at the time of reviewing

7 of 7 people found this review helpful

As the quarter wore on it became painfully clear that the focus on students was minimal for this class. Lectures aligned poorly with homework material, coding assignments were rarely well designed, and grading procedures were at best illogical and at worst completely incomprehensible."

------------------------------------------------------------

"2/5 Stars

Autumn 2011-2012

Sebastian B Thrun

A-

4 of 4 people found this review helpful

This class was a waste of time. Seriously. The only beneficial part to me was the final project. Everything else was so frustratingly simplified and easy that I wanted to slap myself for taking this class.

Here's an actual problem from the midterm that demonstrates our professor's opinion of Stanford students:

For a coin X, we know P(heads) = 0.3 What is P(tails)?

And whenever there was anything REMOTELY difficult, the teachers would, without fail, give a hint...

Just skip to 229. It may be tougher, but this class is not worth it anymore."

------------------------------------------------------------

"User had not rated this course at the time of reviewing

6 of 6 people found this review helpful

Hands down the worst class I have ever taken in my life. This was a joke of a class, far too easy so the curve was mind boggling (the average on the midterm was around 97% because they gave the same test to us as they did to the online class).

Essentially, this class catered to its free online constituency that doesn't pay for tuition. Seriously, I am completely ashamed of this class, and it has no place in one of the best AI universities int he world. Only take it if u need it, otherwise go straight to 229 or something else much better."

------------------------------------------------------------

"User had not rated this course at the time of reviewing

4 of 4 people found this review helpful

This course is useless and you should probably take it only if you have to.

The class should be the Stanford class given to anyone for free, but it has became the class for anyone given at Stanford (where we have to pay for it).

The homeworks were really easy, the only difficulty was to understand what was expected with poor indications. The real lectures sometimes conflicted with the online videos, and in this case the teachers considered the online video as the reference. This means that going to the class every morning instead of looking at free online video gives you a disadvange in this class. (confirmed by the TAs and posted on the class forums).

In one sentence : if possible take the online free class instead of paying for it. It will even be better."

Jeff Atwood’s blog post “Please Don’t Learn to Code” (http://www.codinghorror.com/blog/2012/05/please-dont-learn-t...).

Kurt Vonnegut’s “Harrison Bergeron” from Welcome to the Monkey House

Edit: Fixed whitespace

TL;DR: Learn AI Hard. Thrun make class dumb. Now anybody learn AI.

I'm very surprised that it sounds like he used his Udacity course as the material for the actual Stanford course. That's a huge disservice to Stanford students.

Online education is going to take time to develop. If the courses are too difficult, very few people would make it through them. At this stage, they need the idea to catch on more than anything else. Once it catches on and they can establish legitimacy (i.e. being able to hand out degrees or find people jobs), then it'll be easier to convince prospective students to deal with more challenging material. It's a balancing act.

"If the courses are too difficult, very few people would make it through them" - There's an easy way to solve this. Just have primer courses, or introduction courses which then lead on to hard courses.

I think online is capable of delivering hard courses, not to is giving it a disservice.

I've done several of these online courses so far.

The Udacity ones are definitely the easiest of the bunch. I think they aim to be. That's ok, it seems to be their market, at least at the moment.

Cousera courses seem to be more challenging, though I suspect they still fall somewhat short of the demands of the actual Stanford classes. But I get more out of them and prefer them to the Udacity classes.

The best class, by far, was the MITx Circuits class. I found it very demanding and felt a genuine sense of accomplishment when I completed it. Again, I'm sure it wasn't as tough as the actual MIT class since the exams were longer and open book, but it was by no stretch of the imagination easy.

If all the classes at the various options were like the ones that have completed thus far, I'd pick the EDx (formerly MITx) courses hands down.

That's a good point--Udacity courses are 7 weeks--it seems unfair to compare them to a semester long course.

So much negativity here. Udacity is great. I have a finance background and struggled through some MIT OCW courses when I was learning how to program, and then ended up taking a year's worth of computer science courses at UF. One year later I have my first b2b web application with several dozen paying customers. I'm now taking some udacity courses and boy, I sure wish these existed 2 years ago. Udacity won't turn you into a world class computer scientist, but it is a wonderful way to learn and improve.

"Udacity won't turn you into a world class computer scientist, but it is a wonderful way to learn and improve."

Exactly, Udacity is essentially a more interactive version of w3schools and other tutorial websites. With a more expansive collection of areas of study.

The problem stems from this statement, the crux of this post: "Is Sebastian Thrun's Udacity the future of higher education?"

That's where the negativity comes in. As you said, Udacity won't make you a world class computer scientists, but that's what universities like Stanford, MIT, UW, Michigan, Caltech, CMU and other top CS schools are /supposed/ to do.

If you want to learn what AI is generally about, Udacity can help. But if you want to build a career out it, and become and actual expert, these sites fall flat on their faces. By embracing Udacity and Coursera as tools of higher education, Stanford and universities that follow suit are damaging the quality of the education they provide.

What a complete lack of imagination. You sound like someone who saw the first flight and said "pfft... 100 meters? I can do better on a bicycle".

Think about possibilities.

Imagine what can happen 10 years from now, when Udacity covers the whole curriculum of all major majors, not the few courses that they managed to build in few months of their existence so far.

Imagine what happens if they have 10 years to tweak and improve their lectures, based on feedback on data they gather from past students. They can only get better! (which is not the case in physical universities, due to rotation of lecturers and the fact that some of them were never good to begin with).

Imagine that they hire faculty to start doing real research project, the way MIT, Standford et al do, all coordinated via internet, live video calls etc. Imagine they do it so well that they are allowed to start giving Ph.D.s.

Imagine that they start giving master degrees via testing centers, after you pay a modest fee for taking the test (something they have already started doing).

Imagine that they start coordinating in-person study groups via meetup or some other such service, the way e.g. programmers self-organize and create "Android SF user group" and such.

Those are just 5 minutes ideas that I'm sure are not escaping Thrun - he's much smarter than I am.

The disruption here is zero cost. If they can maintain that and expand to offer more, better courses, it'll be massive.

Your analogy for the current argument is weak at best. A better one would be that:

I saw the first flight and then Wilbur and Orville went to my local railroad magnate and somehow convinced him to destroy half his trains and use their primitive plane for mass transport instead when I had already paid the train magnate most of my savings for a four year contract to haul my goods to a distant city. I'm already upset but I give the two people the benefit of the doubt and send my goods on their plane anyway because there's no alternative. It promptly crashes and burns and I lose all my money.

Are you arguing that Udacity in its current state will make you a world class computer scientist? Because that's what /I/ am arguing against. It is not currently anywhere near a replacement for the current higher education at Stanford. Look at the reviews above for evidence of this.

Edit: Sarcastic response to above commenter's statement about my imagination removed. Downvotes duly noted.

But if you want to build a career out it, and become and actual expert, these sites fall flat on their faces.

I don't think anyone believes that eight days spent learning how to build a driverless car will make you an expert in AI, but what if you followed up with similar, and increasingly challenging, courses for several hundred days like a traditional student would? Where then would you stand in relation to those who went to a traditional institution?

In other words: Is the problem the delivery mechanism or the lack of content, which should improve with time?

I agree, with time the amount of content could drastically increase and Udacity could become a more interactive, education oriented version of Wikipedia. Which would be amazing. Imagine tutorials for everything you could ever dream of!

But I also think it lacks mechanisms comparable to those found in higher education when it comes to fostering students problem solving and critical thinking skills.

One of the biggest challenges when it comes to teaching is figuring out how to help a student solve a problem without giving them the answer outright. Having been a TA at Stanford for two years now, I can say that I have never helped two students work through the same problem in the exact same way.

I fear that with Udacity, users will often reach points where they are stuck, even after hours of trying, and instead of being taught, they will receive the answer.

I can see it already happening in the forums and wikis there. Students can't solve a problem, and they get a solution posted for them. The value of in person education comes from having someone poke you and prod you just enough so that you figure out the problem, but not so much that you can't honestly say the answer wasn't given to you.

If Udacity can figure out how to do that, then I believe that it would rival any higher education system.

Great comment.

A question: in regular universities, students form learning groups. does the kind of learning you describe(somebody teaches somebody) happen inside those groups, or mostly happen only through TA's ?

>>If you want to learn what AI is generally about, Udacity can help. But if you want to build a career out it, and become and actual expert,

Isn't udacity and coursera aimed at college level education? you usually don't become a machine learning expert after college education, but you can have a career developing software(and maybe use machine learning). Maybe not a career at google but still a career.

Yes it is, and you're absolutely right. Nobody is an expert straight out of school. But school lays the foundation for future learning, and if you look at the reviews posted, almost all the posts talk about how easy the class was. It didn't challenge the students. It didn't push them. It didn't teach them. If the foundation isn't solid, then how can you build anything on it?

Remember, the people giving the reviews are stanford students. They are much better academically, and are used to much tougher courses than you're average college student.

On the other hand, i've seen a review from somebody in kenya i think, that said that those online courses are much much better than university courses offered in his country.

Also there are varying levels of difficulty at coursera,MITx and udacity. Someone in this thread said MITx course was challenging. So we still have to see if courses are challenging and to whom.

To address that statement, I'd say it's quite fair to assume that Udacity type model is the future of higher education. Of course, not Udacity as it stands now, but I don't think it's even 18 months old.

I'm sure they'll see the response, and adjust their approach to optimize everything.

The AI class experiment wasn't even a year ago. Udacity as its own entity is barely 6 months old.

What portion of any of the hot online education projects could have been done without the web? For instance, how much of each of them could have been replicated by mailing people DVD/VHS lecture videos and quiz software? If most of it could be done without the web, why wasn't it? Cost? Excessive waiting and friction caused by going through the mail?

I agree with almost everything I've read of what Khan has said, and Thrun is certainly impressive. But when other people go nuts about the possibilities of online education, they seem to imply that the primary hindrance to educational success worldwide has been a lack of access to video lectures. For most students, I do not think that is the case.

The biggest disruption is zero cost, better lecture format and systematic approach.

1. Zero cost.

Yes, there are plenty of web-accessible, paid educational material. Many people make good business from it (e.g. lynda.com). Zero cost is a differentiator.

Also, the paid material is usually vocational (e.g. teaching you how to use photoshop), not physics 101, the kind of material you study in college. I assume that's because college already exists, gives out diplomas so for-profit enterprises don't try to compete with colleges.

2. Better lecture format.

There are free, high-quality lectures (MIT's OCW etc.) but while the content might be high-quality, the experience is awful. At first I was excited about OCW but just couldn't bear to sit through 1 hr lecture with paltry written notes.

Udacity uses the format pioneered by Khan Academy of short videos, has supplemental like exercises and has a clear path from start to finish.

3. Systematic approach.

Again, compared to previous initiatives like MIT's OCW, Udacity's goal is to provide complete courses. Clearly they are at the beginning of delivering that but I think that we can all agree that their Minimal Viable Product has been very successful, which validates product-market fit, and it's also clear from interviews that Udacity has much bigger goals and they are executing pretty aggressively on them. Remember, Facebook didn't have 600 million users on the first day.

I think you've missed another very important part out.

There's a very large, and extremely active community based around each course. Some are necessarily more active than others. Most of them have the actual teachers participating and I've witnessed the forum evolves into a peer-driven help system. Peter Norvig's activity on his course's forum was astounding and I've never seen a more helpful and active online community before.

I haven't seen that with other online courses.

OCW had nothing in terms of community engagement. Coursera does have a forum for each course, but they are very dry and the engagement factor feels missing.

I haven't had a chance to take an offering from MITx yet.

I have seen a few interviews where Udacity staff or teachers have said they deliberately wanted to engage the students and to encourage them to become active and support each other and I've seen that it's quite a big boost to the rest of the offering.

The timing is probably more important than the actual technology used. People's attitudes towards education have changed dramatically in the past few years, in my opinion.

It wasn't long ago that Google bragged that they would only hire the best students from top universities. Now they have hired people just because they have successfully completed a Udacity course, have college dropouts amongst their ranks, you name it. That is a pretty dramatic shift in such a short time, and they're not the only company to follow in those footsteps.

It is not that mail-order education was impossible years ago, but nobody would have taken it seriously. There was a strong leaning that a college degree from a traditional school was the only way to make it. Without that, you were going to flipping burgers at McDonalds.

I think people are now ready for new ways of learning and it just happens that the web is now a great way to distribute that education.

> For most students, I do not think that is the case.

Unfortunately, most students in this world have shamefully little access to quality education. Access is definitely a problem.

Even well-funded education systems have limited access to the best possible resources. The best resources are scattered everywhere, from Stanford professors to worksheets on some teacher's desk in Idaho. Getting them online is a big step.

I think that "consuming education" via video + interactive quizzes (and course-specific chat room & in person meetups) is much more attractive, and engaging, than sitting on your couch watching a DVD.

It is pretty incredible to know that learning a topic of interest is literally a few clicks away...

Also, on another note, it could be that since my cost as a student is almost zero (let's take the value of my time out of the equation) then I am more willing to "try" a class and, since the Udacity classes tend to be engaging, I actually stick around.

Edit: IMO Udacity is a great way to start learning something new, such as basic CS stuff.

At the very least Udacity seems to be winning the PR war. Coursera had the same origin as Thrun's AI class, and arguably had the better platform/user-experience from day one and today offers many more courses than Udacity - yet all the attention seems to be on Udacity only.

I think online, low-cost education will supplant courses taught with low or no instructor-to-student interaction in a very short window of time.

Less obviously, I think that mentors (teachers and professors at all levels) who provide value to individual learners will stay in high demand, along with the institutions that employ them. I also think that the personal networks of these mentors will become the gateway to top-level employment in many fields.

As someone who is vaguely familiar with programming and got about 1/2 way through the Udacity Python course I was disappointed with the education I received. Udacity courses seem to gamify to the extreme, giving hyper-specific tasks and immediate feedback without that much context or exploration.

I decided to sign up for an O'Reilly School of Technology course instead. Yes it's not free, but I fee like I am actually learning something.

I don't believe that the intent is to replace higher education for the tiny fraction of us who are lucky enough to be able to afford it thanks to money, time or location, but to make it much more widely available. Sebastian's quote from the end of the article: "It's the beginning of higher education for everybody."

As a current Udacity Student I believe that we are reaching (or have reached) the tipping point for online education.

The Udacity classes are very informative and relatively engaging (love the lecture quizzes), however, I do believe that there is still a lot of room for improvements.

On another note, I am pretty excited about the upcoming local Udacity Meetups (http://udacity.meetup.com/).

(note: did not read the CNN article) I think that all the complaining about the lack of difficulty is completely missing Thrun's and Norvig's objectives. The most important thing educators provide is inspiration, not information or putting students through mental push ups ad nauseam. I think Thrun's classes are extremely effective in that regard.

Often we forget, that "higher education" is just as much about building networks and adaptation, as learning traditional material. You learn what's acceptable, what kinds of people are successful, and what kinds of people you work best with.

You meet the people who are going to find you that dream job. If you're an entrepreneur, you find the people who you'll want working for/with you. You meet the people who'll help you get things done.

The question here really shouldn't be about whether Udacity is a good educational model. It should be about how far it can go - and I absolutely do not believe it is a replacement for traditional "higher education".

I read somewhere that Udacity's long-term business model is to profit through building those networks. Of course business models can change, but as it stands they have a lot of incentive to provide those foundations, more so than the education itself.

A little off topic, but when is the testing center partnership being started? I read the blog post announcing it, but there was no exact date on when one go and take the test.

Udemy is closer than Udacity to being the future of education but both are far too narrow.

A perfect example of Betteridge's Law of Headlines (http://en.wikipedia.org/wiki/Betteridge%27s_Law_of_Headlines):

“Any headline which ends in a question mark can be answered by the word 'no'”

No, it's a good attempt though. Most of his courses are too theoretical to matter or get anyone anywhere a job.

The exact opposite criticism from me.

Courses should not be "watered down". They should be as hard as they are at top universities, other wise they will lose credibility.

Do primer courses, leading on to the harder courses if you have to.

This revolution should be used to make high-end knowledge available to everyone who seeks it, not to water-down that knowledge. I want online courses, but good ones which are comparable to top courses.

Sorry, I meant they are too niche.

Reply on news.ycombinator.com