|
After I saw first week lesson from Udacity CS101 and SaaS I've only one thing to say: The difference between Coursera and Udacity it's the difference between put a university course on-line and make an on-line university course. Any opinion ? P.S.: and maybe it's the reason whay Sebastian left Stanford/Coursera. |
|
@AINeko: I agree. Re @jholyhead : about Coursera. IMHO it has nothing to do with going broad (Udacity is preparing a lot of new classes for next semester). It has to do with their chosen model for online education. And they are not audacious enough, compared to Udacity, about changing the model of online class. You are right, content IS king, and that is where Udacity shines. Because their content of courses is tailored to an online-class, not just "let's record our lectures, cut them in smaller chunks and throw in questions that check if the student is not asleep". The only Coursera instructor that gets the need to change approach how the course materials are organized seems to be instructor of Model Thinking - Scott E Page. What I saw from SaaS and Crypto did not impress me as a good onlinization (weird noun! :-)) of a normal class. And the argument "they are just starting, it will get better" doesn't cut it. Udacity had just AI class to learn from, Coursera had ML and DB classes. Thrun & Co understood that online class works better with a different model than just a bit shorter University style lectures with PPT slides. Coursera platform is too rigid for that, and most of instructors that jumped the wagon and are providing classes don't really understand the differences and potential, and attached need to change the mindset how they teach. SAAS is clearly just the lecture theatre presentation dubbed over some slides. It's not great from a production quality point of view, but the content is pretty solid and that is what I'm there for. Sebastian' pen is awesome and from the first half dozen videos, it looks and feels awesome and familiar at the same time. I love what they are doing with the code editor. I just think we need to wait a while before judging them. Coursera has a lot of instructors who will take different approaches. Over time, the ones that work best will come to the fore. It's not always nice being the Guinea Pigs. 4
If my only interest was the solid content, I would read their book about SaaS without following the class. From my point of view, the real added value is providing engaging content and the tools to support a vibrant online community. If self-motivation was not an issue, MIT OpenCourseWare would be a lot more popular and there would be less interest in the current developments of on-line teaching. So far, Udacity's CS373 is (for me) way better than Coursera's SaaS. The videos are more dynamic and have a better production quality, the quizzes are innovative and the forum is less clunky. I really hope that the assignments for the SaaS class will raise the bar as I believe that the VM experiments and auto-graded submissions can be a huge motivator, if done correctly. I do agree that grading and tracking of completion are good motivators. But MIT OCW's larger problem was that it's a mix of classes with varying but always substantial amounts of missing content. The stanford openclassroom is a better example of no frills content. As a matter of fact "onlinization" is a noun. It's not the verbing that weirds language, it's the renounification. @sagittarian oops, linguistics has never been my strong side, I just wanted to make a point in a bit funny way |
|
At the moment I think the ml-class style programming assignments where better. Maybe the difficulty will rise, especially with the homework, but the Robotic car Python snippets where extremely trivial, if you are fluent in Python. The problem with easy programming assignments is that they don't make you grow, and you might as well not do them. But as I said, I hope the class is getting more difficult. I didn't get a good score on AI btw (below 90), and still thought their exercises and questions where too easy. Maybe I am a coding masochistic... 3
Plenty of the Machine Learning programming assignments were pretty trivial, in my opinion. When I mentioned it in the forum (something like "no challenge, because too much information and step-by-step walkthrough given in the description") I received a storm of protest and downvotes as a reply ;) I think it's hard for the teachers to adapt to the right level of complexity for so many students with different background, e.g. if you have somebody experienced in building robots who never wrote a line of python before on the one hand, and some experienced programmer whose first contact with AI and robotics was Intro to AI class on the other hand. 1
What I wanted to say is: It'll hardly be possible to offer the right challenge for everybody in a class, but the net (and life) is full of challenges you can choose from. Go, get yourself an Arduino board, connect it to the Washing Machine, and have it move through your flat by controlling its spin cycle ;) I agree with your comment on ml-class's assignments, at times they were not challenging enough for my taste, but at least they required some free-form programming and problem solving. The problem with this not that there is a lack of opportunity for exploration, but most of the time you really need a guide to do meaningful exploration instead of wasting your time. 1
IIRC in ML they decreased the difficulty and increased the walkthroughs after the initial assigment(s?). I think it was at least partly reactive to feedback. @BayesianHorse I thought the (Stanford/Berkeley) AI class pacman assignments were better if you are looking for more challenge. I am hoping the CS373 assignments will build (this is the first week) and become more sophisticated. I am not terribly optimistic though. I think the reality of the situation is that most people do not want a real college level class (much less Stanford) and Udacity/Coursera will need to target the majority. I hope I am wrong. On another note, I like the CS373 in-window python interpreter for simple assignments, but worry how it will scale. I LIKE having a debugger available if I am working on non-trivial code. 1
I'd really like to take a look at their sandboxing technology. Last time I checked nobody dared using a Python sandbox to let untrusted people run code on a server... But nowadays with vms inside vms inside cloud datacenters maybe that changed. 1
@rseiter - "I LIKE having a debugger available if I am working on non-trivial code." A little cut-and-paste into the Python Execution Visualizer and you're in business. Already used it for CS373 (turned out to be an indentation issue). http://www.aiqus.com/questions/37412/resources-python-execution-visualizer @robrambusch that a good idea (it is even nicer than a debugger in some ways). I'm currently using Pydev/Eclipse to play with the CS373 assignments when I feel the need/desire. I guess one point of my comment was that if they expect people to use their environment I'm not sure they can make the assignments very difficult and expect students to have success debugging them. If ml-class is trivial, then CS101 will bore you to death (the presentation is excellent however) @prusswan: Well, yes, but I expected that, since I'm not exactly 101 level. I'm watching some of the videos, because I'm interested in how David Evans teaches the class, but I'd really like to have a 3x speed button @rhasarub, I'm taking CS101 for the same reason. David Evans has a CS textbook out which has good reviews on Amazon http://www.amazon.com/Introduction-Computing-Explorations-Language-Machines/dp/1463687478/ref=sr_1_2?ie=UTF8&qid=1329927649&sr=8-2 1
If the programming exercises are so trivial, why have I not yet heard about your self-driving car? Coding is only to express and implement ideas. The challenge is understanding the material. Once you do, coding should be trivial. 1
You don't honestly believe that these exercises are comparable in difficulty to making a self-driving car? It's just a motivational theme, in case you were wondering... 1
Most people aren't and it wasn't a hard requirement for the class. I'm not saying that the questions are hard, but neither were the ML ones (especially if you already knew Matlab/Octave). I think Thrun once said that after teaching a class with 100,000 students (online) he couldn't go back to teaching only a 100 (at Stanford), so I think that he's aiming to include as many people as possible. That means not assuming too much prior knowledge and not making things so hard/challenging that a large number of people are unable to follow. BTW, if you're looking for less hand-holding, I recommend you try the DB and/or SaaS classes (I haven't tried Crypto myself so I don't know about that one). |
|
Thrun left Stanford to work at Google, where he still works to this day. Udacity is taking the model of the AI-Class and refining it. They only have 2 courses at the moment and have been able to spend time working on production quality (I love the new pen). Coursera are trying to go broad. They have a lot of courses, from a lot of instructors at multiple universities. Their focus is on getting as much content out there as possible. And yeah, production values seem to be secondary, but content is king and they will (eventually) have more of it available than Udacity. Over time, hopefully this gap will narrow as they learn from their own failures and from each others successes, with Coursera spending more time on production issues and Udacity having a bigger spread of courses. It's early days. 2
Actually Sebastian left to start Udacity: http://blogs.reuters.com/felix-salmon/2012/01/23/udacity-and-the-future-of-online-universities/ He always worked at google =) 4
"I did on my own volition resign from my full tenured position, effective April 1, 2011. However, this was primarily to continue my employment with Google, and it predates my online classes." crap! =D I read that when it was written and just looked for it again lol sorry... 1
LUMB! The story of Sebastian Thrun leaving Stanford has been an education in how misinformation & rumors propagate, driving out mere facts. He's still associated with Stanford, right? @EllenL - Thanks for lumb. ;-) 1
Well, I was not very clear :) @jholyhead I see that update on the article, but I am confused by it. Firstly, I can't find it anywhere on Thrun's site when I google for "volition site:robot.cc". Secondly, AI-Class started in Oct 2011; had he really left Stanford back in April 2011 and not mentioned it all during AI-Class? Finally, could there be any significance to the date 1 April? |
|
Here's a more statistical answer to this. I hope this data is somewhat reliable, and sorry for putting in all those images, I hope you won't mind: At the moment, both Udacity and Coursera have two classes running since February 20, so in fact it is possible to compare them on a somewhat equal level. According to HQindex, Udacity has 47523 unique visits a days, while Coursera has 24298. Meaning Udacity attracts twice as many students per class. But one has to take into account maybe, that Coursera's Model Thinking class posted all videos for the first two weeks immediately, and there's has not been any new material in the second week. Thus, there has been a drop of traffic in the second week, as you can see on the graphs below (taken from Alexa):
Still, you can see that Udacity's traffic rank has been higher for Udacity so far. Another interesting thing Alexa.com showed was the origin of visitors. While the basic distribution is similar, it is remarkable that you will see far more Russians and Brazilians among the Coursera traffic, which I also noticed in the Coursera forums. Udacity on the other hand has a higher rank in the english-speaking world. Here's the details:
2
One more non-statistical thing I'd like to add: I think Udacity is doing better so far including the students into the process. The Teaching Assistants show very active participation in the forums, and when posting bugs and valid suggestions, there is actually somebody who will take care of this. They seem to have learned from the Aiqus experience. |
|
Slightly off-topic - looking at this video (CS101->Unit 1->Capital Udacity), our guesses about roots of the name Udacity were correct :-) |
|
I haven't seen the other current Coursera classes yet, but from what I've seen, the Model Thinking class by Scott E Page is definitely not a 'put stuff online' course, he manages to make a pretty nice multimedia experience of his lectures. They are less interactive, there is no python shell thingy, but those in the CS373 Udactity course don't add anything in my opinion. In fact, they have the feeling of being there because they want to show off their cool technology, not because it actually helps; those small coding quizzes were extremely trivial 'fill-in' type of things. So far, my vote is for Model Thinking as the current 'king of multimedia lectures' :) 3
I don't agree with your opinion on Udacity courses, I think the Python exercises are cool. Your mileage may vary because of programming (Python) background. I found it very cool to do it "my way" and then look at alternative method. I agree on Model Thinking as a very good class. I did however catch myself couple of times in the MT lectures on a thought "there should be a quiz here to test for intuition and reinforce learning". 4
@Gundega -
I find that happens to me in conversations with fellow humans now. When someone explains something to me and rambles on for a few minutes, I find myself thinking "All well and good but this conversation needs a quiz!!". ;-) @robrambusch -- that's a sure sign that you need to get out more ;-0 3
What I think is Model THinking = Udacity >>>> SaaS Model Thinking and Udacity are way much better than SaaS course.. But we are two full weeks into the MT class and there are still no assignments or quizzes and no new lectures have been added since day one. I'm wondering if the Prof. may have dropped the class. I don't think he dropped the course. He already produced 2 weeks contents and displayed it on website. It might be under working.Anyway This doesn't matter much .. Because from Tomorrow morning MITx course is starting..& RELATE is also good. Computer Vision will start too.. I am not sure about Stanford courses anyway... |
|
Quite an astute observation =) I have to admit that for me it isn't much of a problem. I decided to travel rather than going to uni and I am now 30 with a 10 month old son and I have neither the time nor the resources to go to uni, so I am willing to over look small problems as what I am getting in return is well worth it. But I can understand that some people might be disappointed. |
|
I don't know when Coursera's stanford courses will start.. but I think Udacity oncoming courses(8 courses) look cool. I'm joining them all if Coursera fail to start its courses until April Mid.. How many are waiting for Coursera's courses for last 2 months like me? I've asked query about When udacity courses are starting ? I got reply that was.. Hi XXXX, When other Coming Soon classes will come? The courses will be offered again starting April 16, 2012. Thank you for your interest in Udacity and we hope you enjoy our courses! Thanks Following courses will start.. 1.THEORY OF COMPUTATION 2.OPERATING SYSTEMS 3.COMPUTER NETWORKS 4.DISTRIBUTED SYSTEMS 5.COMPUTER SECURITY 6.ALGORITHMS AND DATA STRUCTURES 7.SOFTWARE ENGINEERING PRACTICES 8.BUILDING WEB APPLICATIONS It looks like whole 1 semester of computer engineering course.. |
|
RIght now, I would rate Udacity & MT in the same level. SAAS looks like "just put the course online". |
|
And now there's a new kid on the block. The first MITx course, 6.002x is up and running, and so far, to my eyes, it looks way better than last year's AI & ML classes, or Udacity's current ones. I'm not doing SaaS or MT, so can't compare, but from comments here it doesn't sound like either is any great leap over last year's inaugural classes. 6.002x looks like it is. Well worth checking out, even if the topic (Circuits & Electronics) is not of interest. |
|
My experience is the reverse of that of the OP. I am enrolled in the Coursera Compilers course, which is a bit advanced for me but I am thoroughly enjoying the ride, and am enrolled in the Udacity CS101 and CS212 courses. My background in programming is mainly with C and to a lesser extent C# and I have no experience with Python. The Udacity courses are slick but so far appear superficial while Coursera offers a solid, highly educational course. Yes the Coursera presentation is less slick but that is a minor issue. CS212 does offer good material for someone starting programming, you just have to know python, so it depends to some extent on your background. |
|
Off topic, Did someone here have been asked to give feedback on pre-launching videos of Udacity?? Yes, I was asked to give feedback. This version is a considerable improvement over the original. Like day and night (or vice versa) @jimgb depends on what part of the diurnal cycle you enjoy best i suppose =) |
|
Hey all, Not sure if this is the right place to post this but I've taken a few courses from Udacity and Coursera and I find that it is difficult to meet up with others who are taking the same course. Want to gather some feedback if meeting up with others who are taking the same course around you is important to you? http://signupstudy.ninebricks.com/ Thx! |
or 
