Hacker News Reader

I very strongly agree. Udacity courses gives you this feeling that they step back, and tried to reinvent how would an online course be if it was invented today, taking advantage of the tools we have today. While coursera has that feeling that they took existing courses, and just threw it online and hoped for the best. For example, I have the feeling that Python was chosen for udacity (among other reasons) specifically because it could have an online interpreter. While coursera feels like they just reused whatever language and tools they were using before, just because. If I try to consolidate the udacity advantages it would be:

- Short videos: This have many different advantages and works well if its other properties. Let's you set your own pace, instead of following the long video's pace. Easier to separate topics into different parts you can easily click to go back to. When necessary, it's easier to pause between topics (since the video pauses automatically) for you to take a step back and think about you just learned for a minute. It matches the interpreter nicely, learn something new, try it, learn something else, try it. It allows the instructor to better separate links to appropriate material in the description, by topic.

- Online interpreter and automated tools: I suppose this is one of the reasons why they chose python. Similar to how codeacademy works, udacity is like codeacademy + video instructions instead of text. And this is genius. The videos are taking into account that you have an interpreter with you, and everything is integrated. The lessons, quizes and homeworks are all automatically integrated with the interpreter to make sure that you try out what's being taught and the class can give you instant feedback if you're doing it right.

- Production value: Not only the videos look better, but they are edited to be more efficient. Unnecessary parts are cut out for brevity. The video and sound quality makes sure everything is easy to see and hear. Some other online courses often times display a printscreen of the IDE, with low resolution, which makes the code unreadable. On udacity, the code is always readable and the audio always easy to listen. The better UI makes it easier to navigate. And of course, the superior visual quality just makes the whole thing more pleasant to consume.

These 3 characteristics I wish every online course would try to learn from udacity and implement themselves. It's important to note how these 3 are glued together and help each other. Short videos gives you breaks to use the interpreter. Higher visual quality makes the code easier to read and then try on the interpreter, etc etc Udacity just feels like a consistent and smart bunch of ideas put together to make sure we learn easier. Hopefully, others will evolve with them.

It's barely 6 months into this whole thing, and Udacity and Coursera are younger even than that. The first gen videos for Thrun's AI class were awful. Change has already happened fast, and there's plenty more to come.

I wonder if Coursera's great number of professors and classes will slow them down in offering a drastically different format. Udacity has been exclusively Python (with snippets of HTML in the web class), and their IDE is shared across all the classes. Coursera's many languages and professors might make that degree of focus much more difficult.

Agree, the first AI class was horrible, while Andrew Ng's Machine Learning class on Coursera was top notch. I think both platforms have advantages and things to improve, and I'm glad they are both activelly working on it.

Not sure we need yet another platform with eDX, but let's see. So far it seems more of a West versus East type of thing..

Well, I prefer Video Lectures (Like the MIT Python one) to Udacity and Coursera, feels more real and I understand it better. Also I prefer long videos (About 1 hour) to that 5 minutes videos from Udacity.

I gave up in both Udacity and Coursera (Though, I liked Coursera more), and still love those MIT Video Lectures.

I think it's a matter of preference.

I liked Udacity best at first, but after a while found myself spending a lot more time with Coursera. Mainly for two reasons:

- Udacity sometimes belabors some really basic points. Coursera never gives me that "dumbed-down" feeling. It's more like a real, and challenging, college course.

- Coursera lets me speed up the lecture, as much as 2x, and every time I go to a new lecture it remembers that speed. I managed to speed up Udacity lectures by going to html5 in youtube, but I had to reset it every time I started a new 2-minute segment.

"Udacity sometimes belabors some really basic points. Coursera never gives me that 'dumbed-down' feeling. It's more like a real, and challenging, college course."

I think this actually may be Udacity's biggest strength. It's a lot more hands on than a college course, which makes it somewhat easier. But do you actually learn less? After thinking about it, I really don't think so.

With your average college CS class they spend a few minutes talking about some problem, and then you spend a week on your own working on it. Whereas with Udacity they actually teach you all the concepts you need to solve the problem, and then you work through it together step by step.

The strength of the traditional college approach is that it teaches you to be resourceful to some extent. But this comes at an enormous cost. You basically spend all week writing shitty code and trying to make it compile just to learn at best one or two concepts. Whereas with Udacity they are actually teaching you how to write quality code, but you still have to do the most challenging parts of the algorithm yourself. (And these aren't trivial, they often take several hours.)

Clearly at some point you do need to learn to fly yourself, but with Udacity at least you can wait until you're up to speed on the basic tools rather than being shoved out of the nest on day one. Even if it seems dumbed down, I'm completely convinced that you're actually learning much more this way than what you learn in the traditional CS classroom. And after all, the measure of a class shouldn't be how hard or easy it is, but rather how much you learn from it.

I like the short cycles and quick bits of programming, especially given that they've also got the weekly homeworks. I'm just saying that sometimes they move too slowly for my taste. Let me run the videos at high speed without hassle, and that'll help. I can slow down when I need to, and speed through the parts I already get. (I've found that at high speed, having to concentrate a bit just to understand the words helps keep me focused.)

Ultimately, maybe we'll end up with classes that have several levels of explanation, letting you pick which you think best suits you for that particular piece...maybe for one part a quick presentation of an equation is sufficient, and for a less familiar part you want a step-by-step walkthrough with lots of intuitive explanation.

Hmm. Telescopic lectures ala http://www.telescopictext.com/ .. You get a short explanation, some tests, and if you get below a certain score, they go into a more detailed explanation.

Being able to speed up the lectures is a huge nice-to-have.

You do know you can speed things up in vlc? VLC also does a great job of keeping the pitch accurate.

Are you saying you're able to watch the Udacity videos in VLC? How does that work?

It is quite easy. The braindead simple way is to give VLC the url for the video. But my preferred way is to use youtube-dl to download the videos and then open the video in VLC.

I'm going through the Coursera machine learning class right now and I have to say that the professor glosses over several details and often makes comments like "if you're not familiar with calculus..." and "if you're not familiar with statistics..." which caught me off guard at first. I really doubt that actual Stanford students enrolled in a machine learning course would be lost on the incredibly basic operations (e.g., taking the partial derivative of a polynomial function) he is using.

Also, there has been no acknowledgement of how contrived the exercises are. For instance: exercise one gives a data set of a the profitability of a company's existing stores versus the population size of the city in which the store is located (in units of 10,000 dollars and people, respectively). The range of the data is 5-23 (population), with most of it concentrated below 10. We fit a straight line to the data using least squares, then use that line to predict the profitability of two new locations--in cities of populations 35 and 75. I understand that this is an intro course, but there is not a word about how ridiculous this is.

I don't mean to be overly negative. I am enjoying the course, but I am surprised a bit by how basic it is. Let me say that I do like the approach of the course to ML, which is to formulate a parameterized cost function and then minimize it by some general method, rather than the typical statistics course approach which is to solve ordinary least squares directly, which gives an "exact solution" (given the data) but does not generalize to more general models.

I know this is foundational material and overall, I am impressed by the approach of the course, but I would expect more comments on the weakness of the naïve methods we are employing at this early stage and how they will eventually be improved. I find it very helpful when professors at least reference more advanced methods or provide references for further reading by the interested student. Admittedly, that is more frequently a feature of graduate courses, but encouraging students to go beyond the material is an important aspect of good teaching. I have watched the videos for several other online courses and I do appreciate the fact that Coursera is allowing me to hand in assignments for grading, which vastly increases by engagement with the material. This, in fact, is the most valuable resource offered by the program. The lectures themselves are fine--if a bit dry--but a good book or a set of well-prepared notes (not slides) would probably suffice just as well if accompanied by the assignment grader.

All in all, this is great. The more people who know about machine learning (and have access to higher education in general), the better.

I'm pretty rusty on my math, so I guess it happens to hit the sweet spot for me at the moment. Once I get back up to speed on calculus I might feel different about it.

You might be interested in this version at CalTech: http://work.caltech.edu/telecourse.html

I just became aware of the CalTech course. I will likely check it out, although I will stay with the Stanford course for the time being.

IIRC from a Stanford student's comment on HN, Stanford offers two versions of machine learning, one that is more math focused and a more applied one designed for all majors. The ML course offered through Udacity is the latter one.

Completely agree. I'm currently taking 4 Udacity couses, and they are phenomenal. There is no question, in my mind at least, that Udacity or at least the Udacity approach is the future of the space.

Their advantage isn't just their technology platform, it's that they seem to have impeccable taste and judgment. They aren't just a bunch of MBAs who are 'licensing content' as part of an 'online distribution play' or whatever.

I actually think that Udacity is quite similar to YC circa 2005/2006 in a lot of respects.

I think there's a fair bit of variation between the Coursera courses.

I'm currently taking their Machine Learning and Computer Vision courses. The ML course is quite like what you described for Udacity: there are a lot of short lecture videos with questions embedded in them, online quizzes and automatically graded programming assignments. It's a very engaging way to learn and I'm getting a lot out of it. The CV course, on the other hand, has so far just been videos of the professor talking & no supplementary material.

I suspect that Coursera provides the platform and leaves how the material is presented on it up to the professor. The ML course has been run once already, so they presumably have the benefit of experience; whereas it's the first time the CV course has been run so it's quite possible that they just haven't figured out how to take full advantage of the platform yet.

>"It's a very engaging way to learn and I'm getting a lot out of it. The CV course, on the other hand, has so far just been videos of the professor talking & no supplementary material. I suspect that Coursera provides the platform and leaves how the material is presented on it up to the professor."

It's probably this. I'm taking Model Thinking on Coursera and there are at least 2 pdf for each section (20 sections). Some of them from professor's book, some links on the web or some scientific articles.

I even stop reading it all because it's time consuming, but I'm happy it's been offered anyway.

About the style, well, it fits in the Content: programming stuff is better on Udacity; academical/theorical stuff, Coursera.

I agree strongly with these observations, and have found it to hold true over Udacity's Robotics AI, Web Apps, Programming Languages, and Design courses, and over portions of Coursera's Machine learning and NLP classes.

Edit: and I also agree with Brunov, Udacity courses are not mathematically rigorous.

On the other hand, I find the Coursera classes much better in quality. At least the CS ones seem to have a more formal and rigorous tone, which is something that I value more than presentation.

I agree though, the Udacity lectures are more engaging.

I don't know enough about programming to make a comment on the quality, and I should have mentioned that in my initial post. I was trying to convey my experience of the teaching in each course, not the quality of the material.

But yes, from what I've taken in the AI class so far they are definitely teaching Computer Science. Udacity so far seems to be "applied programming" with a bit of helpful CS here and there.

Have you taken either of the CS101 courses? Udacity's has already wrapped up and started a new 'hexamester' though Coursera's has just started. I prefer the Coursera lectures generally but with CS101 at least I think Udacity was much more engaging and challenging for beginners. So far the assignments on Coursera CS101 assignments feel pretty lackluster compared to the other classes (unless I've become unwittingly jaded, in which case I'd still like to hear what others think).

I've been looking at the courses from both and personally I like the Coursera set up much better - the constant interruption by questions makes it a lot less engaging, imho.

I agree, Udacity is definitely not an unqualified, clear winner. I like the Coursera 8-10-15 minute videos a lot better than the 2-3 minute videos from Udacity.

I've seen examples of both sorts of courses from both Udacity and Coursera, so I think it really just comes down to the extent to which the instructors are willing to reinvent their material. And for those who aren't doing so much reinvention, I'm still glad they're doing what they are - I'd rather have an online class based on recycled lecture videos, than no online class.

I've noticed this, too. Udacity's courses are clearly designed from the start to work well on the web, whereas many Coursera courses seem to be existing lectures broken into smaller chunks. This really sets Udacity apart, but I think it's actually the result of a much bigger advantage: Udacity isn't affiliated with a meatspace university.

It might not seem like an advantage now, since they're competing with more prestigious options from Stanford and MIT, but I think the decision to detach Udacity from Stanford will prove to be a very smart choice. I think we'll see real universities take a two-faced approach to online learning: they will want the attention it draws, but will for fear of diluting the prestige value of their brands. (See the protracted wrangling over the certificates in the original AI course. And can you blame them? The brand is probably the most valuable part of a modern diploma).

This is Udacity's real advantage: they don't have to worry about prerequisites, four-year graduation rates, department budgets, or the physical constraints of offline universities. They are already much more nimble and much less risk-averse than some of the other initiatives, and it will be difficult for the programs still affiliated with universities (institutions hanging on to practices invented in, like, the 13th century) to detach themselves from offline constraints and adapt to the web.

I'm not sure that beating the traditional university is the objective or that it's even necessary. Remember that the essence of a disruptive product is that it's a little bit worse and (at least) a little bit cheaper than the incumbent product. The question, then, for disruptive education plays (of which my company is one) is: which part of the traditional experience are you going to take away in order to achieve scale? Most .edu startups are taking away instructor interaction and emphasizing lectures, which is exactly the opposite of what most students actually need.

That's why I love HN: just discovered Udacity through your comment and onto my first search engine course. Thank you.

You should know that it's more "cs/programming taugh by building a basic search engine"

I just took it and loved it. You do build a search engine, but te real focus is learning programming. It's an intro course.