[king_magic]

To be brutally honest, I've been pretty unimpressed with Udacity - specifically, their Deep Learning and AI for Robotics courses. I personally felt the quality of the content was low enough that I could not justify paying for something larger, like their Self-Driving Car Nanodegree (which I was accepted to for their first run, but I ultimately declined to shell out $2,400 USD because of the quality issues I observed in those other classes).

Compared to some of the courses I've taken on Coursera (e.g. Andrew Ng's Machine Learning), the general quality from Coursera blew Udacity out of the water.

End of the day, I'm not going to pay for content that is poorly stitched together, contradictory, constantly interrupting you, short on delivering insightful explanations and simply unclear in many places. I'm not going to pay to waste my time on forums to tease out information for solving quizzes / programming challenges that should have been covered in the course content. I'm not going to pay to deal with snarky TA's on a power trip. I'm not going to pay to waste my time because the course quizzes or programming assignments are expecting you to magically gain some insight that could not be reasonably attained by viewing the course content only.

And I'm not saying these courses should necessarily be easy - not at all - but there should be some reasonable level of success attainable without having to endlessly scour the internet for why your solution, which looks pretty correct based on the course material, isn't passing in their test harness - only to learn that the author of the course decided to arbitrarily switch the order of two operations in his solution (that were previously demonstrated over and over in the reverse order), and that's why your submission is failing. Sorry, but screw that.

Not worth my time, not worth my money. And to anyone else about to shell out a large chunk of cash to Udacity - think long and hard before you do - there are likely better options out there.

[garysieling]

I've been trying to build out a site of free standalone lectures (https://www.findlectures.com), and it's become clearer as I've been building a collection of content how different the motivations behind free material is from paid.

If a university / professor / hiring manager is behind the free content, they're promoting their brand, so they need really compelling material to stand out from everyone else.

[king_magic]

Yeah, I've definitely felt hints of the course author's motivations behind the various classes I've taken. For me, though, the quality of the free material is a pretty significant measure of paid material - if a site like Udacity can't put out well constructed material, I'm just not going to give them my money to find out if their paid material is much better.

[cr0sh]

Hmm - your experience seems different from mine; but maybe you've taken more courses - or more recently?

My first experience with MOOCs was the Fall 2011 Stanford courses (AI Class and ML Class) - which ultimately spawned Udacity and Coursera (respectively).

I wasn't able to finish the AI Class, but I did complete the ML Class. In the spring of 2012, Udacity announced their "CS373 - How to Build a Self-Driving Vehicle" course, which was supposed to be what Thrun could offer as comparable to the old AI Class (CS373 became the AI for Robotics course, IIRC). Later Udacity was able to offer the AI Class content. Coursera from the beginning offered the ML Class content. In both of these, I don't know how they each compare to the inaugural Stanford courses.

I want to note here that when I say "Stanford", I am not meaning to imply that Stanford offered them, or you got credit or anything like that - it was just that these courses were initially linked to them, via the instructors and the "experiment" in MOOCs. The response was so large, that the spinoff of Udacity and Coursera was the result.

Anyhow - I found that the CS373 course was really tough for me, though I was able to complete it fully. For me, it really help to open my eyes and mind more on how certain things worked (Kalman filters, PID, etc), and expanded on things I learned in the ML Class (ANNs especially). It also highlighted areas I needed help with (probability and stats, mainly). Coding wasn't the issue, as I has been employed as a software developer for over 20 years.

When the Self-Driving Car Nanodegree popped up, I jumped at the chance. I got in, I paid my initial money. I guess we'll see if it is worth it. I have no illusions that I am going to "land a new job" from this - if I do, great. I do hope that it will further my knowledge and understanding though in this field, and maybe it can help me with other things (I dabble in hobbyist robotics, for instance).

So yeah - one could say I am spending $2400.00 on a lark, but I have wasted similar large amounts for less on the payback end (worst one was paying for a year at TechShop - which I only went to a few times - but I know how to use a laser cutter and 3d printer now - w00t?). I am trusting that the level of the course will be on par with the CS373 course; I guess we'll see.

[criddell]

Ng's Machine Learning course is an outlier I think. It's easily the best online course I've ever taken.

[esfandia]

Jennifer Widom's database course was part of the first three MOOCs offered in 2011, and I thought it was excellent also.

[criddell]

The other course from Coursera that I thought was outstanding was the Cryptography course by Dan Boneh. It covers a lot of material in a very practical, hand-on way.