Udacity Says It Can Teach Tech Skills to Millions, and Fast | Hacker News Reader

Udacity Says It Can Teach Tech Skills to Millions, and Fast

nytimes.com

37 pointsolalonde47 comments

I prefer the MIT OpenCourseware style, where there's no signing up, no trying to get money out of you and you can come in and just take what you want out of a course, much of it graduate level classes where knowledge is dumped for it's own sake, whereas Udacity/Edx/Coursera all seem geared to introductory courses and selling credentials. I typically just want the lecture notes, assignments, and the recommended reading now. I find video lectures too clunky to load up in the bulky UIs of MOOCs, bandwidth problems, and they eat up too much screen space for working along writing out the programs. I appreciate the condensed versions in the notes too since I don't have a lot of time to dedicate an hour to watch a lecture I could read much faster.

For example this course on Elliptic Curves http://ocw.mit.edu/courses/mathematics/18-783-elliptic-curve... I decided to take after going through djb IETF working group mailing list posts which are so detailed I wanted to understand the basic math involved. Every post he writes is so thorough it's typically almost a class in itself in modern security analysis http://www.ietf.org/mail-archive/web/cfrg/current/msg07335.h...

Also, I originally learned "real development" (my decade of hacking around C willy nilly modifying programs doesn't really count) from Sussman's "Adventures in Advanced Symbolic Programming" where you read chapters out of SICP and then work on assignments, with mini lectures built into them in just plain text http://ocw.mit.edu/courses/electrical-engineering-and-comput...

I agree completely. MOOCs require too much commitment to be frank. I'm currently finishing my EE degree, so I can't really afford to "follow along" given my already full course lod every semester. In addition, I'm more interested in slightly advanced topics covered in a rigorous manner.

MIT OCW is the best solution. The self-paced courses on Coursera and Udacity are OK, but the UI as you say is too clunky, and the content is a bit shallow.

Right now I'm starting to work through Computer Systems Security on OCW [1]. They provide lectures on Youtube, so no need to fiddle around with a third-party app on mobile. They also provide all lecture notes, exams, and labs for learning purposes. The content of the course itself looks extremely interesting as well, and covers a wide array of topics related mainly to RE.

[1]: http://ocw.mit.edu/courses/electrical-engineering-and-comput...

You can also look up the recent course and usually get updated lecture notes too, often course calendars are available at MIT and not locked for student authentication http://css.csail.mit.edu/6.858/2015/schedule.html

Speaking of Prof. Zeldovich he's releasing this traffic analysis resistant messenging scheme with some grad students in October http://people.csail.mit.edu/nickolai/papers/vandenhooff-vuvu...

On the other hand, the commitment to a MOOC's schedule can help people stay focused on the course and see it through even if they lose interest at points.

For example: Coursera's NAND2Tetris class imposed a schedule that kept me on track, even when I 'didn't want' to work on it.

The same thing's happening again with their MOS transistor class [1]. I never had a completely firm grasp on semiconductor basics in undergrad, and now I've got the motivation to go back and read through old notes, correct misunderstandings, and so on.

[1] https://www.coursera.org/learn/mosfet

I would not lump edX in with Coursera and Udacity. As noted in Wikipedia: "EdX differs from other MOOC providers, such as Coursera and Udacity, in that it is a nonprofit organization and runs on open-source software."

At first, it seemed to me like Coursera, Udacity and EdX offered similar quality courses. But at this point, I think that, on average, the EdX ones tend to be significantly better. And that difference does seem to me to be related to the fact that EdX is not a for-profit institution trying to extract money from you. Also keep in mind that the EdX systems were largely developed at MIT, and EdX was originally intended as the improved version of OCW.

I would. Just looking at CS, for example, on OCW you can get the any course needed for a degree (or even some graduate degrees). In contrast, on edX the vast majority of the courses are either introductory, sub-introductory (e.g. Computing: Art, Magic and Science) or an intro to <random MS tech>. Even when filtering down to "advanced" the top result is a freshman level course—Data Structures.

Furthermore OCW is rigorous. It includes actual homework assignments given and final exams from previous classes of MIT students. Coursera and edX are extremely focused on videos, which IMO are not the most time-efficient or thorough way to learn. The one area in which Coursera and edX shine is their automated program graders.

On the whole, I would say that if you want to learn math, CS or any hard science and you hope to actually pass a subject GRE, apply to grad school or build a truly strong foundation, then OCW wins hands down. If you want to pick up a new programming language or tool, especially one from Microsoft, then use edX (or Coursera).

Never heard of Sussman's AASP course, is this a new one ?

It's a course on symbolic programming that builds on 6.034 (AI). The lectures aren't open but the same information is all in the suggested reading as you go through the problem sets. Build things like a compiler for a DSL, pattern matching and decision making programs, experimental parallel concurrency through message passing 'actors', most of it is exploratory programming and lot's of SICP/GNU Scheme documentation reading http://groups.csail.mit.edu/mac/users/gjs/6.945/index.html

Always inspiring topics isn't it.

I'm the Director of Engineering at Udacity and wanted to clarify a couple things: All of our content is available at no cost. It's an important manifestation of one of our most important core values—that education is available to everyone. What we charge for are services. Students enrolled in Nanodegree Programs have access to coaches, a massive worldwide network of code reviewers, certification, job placement and much, much more.

We work hard to help Nanodegree Program students that are looking for jobs in tech get them. For example, one of the best parts of my job is doing mock interviews. Folks on the engineering team volunteer to give a representative tech industry interview to a soon-to-be-graduate and then give them detailed feedback on how they could improve. Hearing back from these students a few weeks later that they got a job at Nest or Google or wherever is one of the most satisfying things I've experienced in my career, and everyone on the team who's done it feels the same.

Finally, we really do think our graduates are ready for jobs. So much so that we've hired two of them as full-time SWEs onto the small engineering team at Udacity. This has worked out incredibly well—these are smart, effective engineers that I'd put up against any traditional school's CS grads. We'll be hiring more of them, and we hear from companies that have hired our graduates that they'll be doing the same.

I'm a little disappointed at the comments here. It's fine to point out flaws and downsides. comparisons to other options like traditional universities, free options. Udacity might even ultimately suck.

But, this is very early in the game and MOOCs are part of the progress, whatever their eventual form or role.

The price of education will ultimately come down to a bunch of things that are still unclear. Can it work at scale without dedicated teachers, dedicated full time pursuit. Can credentialing be solved or bypassed at scale.

The "people want it" is there and the only way to figure things out is to try scaling. That could take decades, culture takes time to evolve.

I don't care about Udacity or anything else, but I'd like to see the concept of better education, cheaper using technology. So, I'm happy to hear them say something like "can be scaled up to teach millions of people technical skills." Try it. If they succeed, great.

I don't think the criticism is against the concept of MOOCs but rather the over-exaggerated claims being made regarding proficiency achieved from the cursory information provided in a given program and career opportunities available after completing said program.

I recently completed 7 Udacity courses in less than 2 months. All of them are from the 'Front End Developer' nanodegree. I am not enrolled in the program, I am just taking the courses for free by watching the lecture notes and doing the short quizzes.

Pros:

1. It provides good breadth of knowledge around what is possible.

2. The style is engaging and the videos are short and well made. I have tried to take Coursera courses before, but always ended up dropping out after the first couple of lectures.

3. It gives you a specific goal; most of the nanodegree courses are geared towards completing a specific project.

4. I was able to quickly get my personal website up and running (I believe in using completely static HTML + CSS, JS when required, and the courses helped me to learn enough to quickly hack together something just using Notepad++).

5. I have plans to enroll for a Nanodegree later. It is quite affordable even for people in developing countries (like me), after accounting for exchange rate fluctuations and cost of living.

Cons:

1. The courses don't go too deep, and you won't become a master unless you are motivated enough to do the work yourself (e.g. Introduction to jQuery - it gives a flavour of what is possible to do with jQuery, but does not go into any depth if you want to be an expert).

2. What you get out of the course if entirely up to your motivation in improving yourself. Completing the courses alone won't be enough to pick up a new skill.

Wait until you have more than one page and want to change your menu. Then you won't believe in completely static HTML + CSS.

I used Jekyll for my blog, but ultimately it's just not worth the effort and definitely impractical.

Off topic, but I use Jekyll for my blog, too.

It's compiled to static HTML, sure, but the templating is pretty powerful (downsides of Liquid aside). It's easy to make global changes with templates and CSS.

Octopress can make a lot of Jekyll tasks easier -- https://github.com/octopress/octopress.

"Lower education costs"? Nah i doubt it. Maybe just for the US, but not for the rest of the world.

At $ 200 per a month it is almost as expensive as a top tier university in South Africa.

Udacity Cost per a Month: $200

Cost at a South African University: +- $3000 per year/10 months = $300 per a month.

This excludes any discounts, bursaries etc. Also ignoring the different levels of quality and depth of what you will be learning.

Are those prices with government subsidies though?

I don't know about South Africa, but for Chile the values are similar. An engineering degree in a top-tier university is around USD 4500K per year, no government subsides. There are subsides and scholarships but the normal price is around USD 375 per month.

You don't need to go that far. Even in europe univs are quite affordable. I attended engineering at Politecnico of Milano and it costed me well south of 3k/year.

It seems it's mostly the US having an issue in education costs.

I think the distinction here is between "cost to provide education" and "direct costs to students." You can, through government subsidy, have an expensive-to-provide education delivered at a low cost to students.

Most public universities around the globe that are cheap to students are cheap because of these subsidies. Udacity, if you believe is providing a comparable good, is going the cheap-to-provide route. Admirable if it works, because we can educate the masses without spending a lot as a society.

As someone who has spent a fair amount of time in academia (masters, first half of phd), I see areas where Udacity is a good substitute, and areas where it is not. If you're an undergrad sitting in the back row of a lecture hall, why not Udacity? If you're a grad student doing research, obviously not. If you're a grad student doing advanced coursework... maybe?

Great point. I think the question is, how do you scale this so for example a University here in South Africa. That may not have the resources is able to tap into the flood of content that is being created in the education space, reducing the cost for students here in South Africa and uplifting millions.

You could see this as the free market in the US responding and allowing viable competition for the product of higher education. The commercial MOOC providers can't compete in Europe against government subsidized education.

It is possible to get actual transferable higher education credits for around 100 euros for some european MOOCs. The US MOOCs offer odd bits of paper at this price range because the US credential market is not free at all. Accreditation provides specific barriers to entry in the US to make education cost enough to satisfy the institutions.

For European systems, any student who is taking MOOCs is cutting their costs instead of threatening their profit model. Many students in Europe wont have financial incentives to use MOOCs for their primary area of study, but the European model doesn't really offer unlimited free higher education in what one wants whenever one wants locally.

> It seems it's mostly the US having an issue in education costs.

Are the prices in the US not relative to the GDP per capita? $3k-$4k USD might be cheap in the US but that money means a lot more in different economies.

That's how much it costs in America. But there's nothing stopping them from selling it to South Africans for $20 a month. The marginal cost of an additional student is surely very low.

After graduating from a nanodegree program, Udacity now reimburses students half the cost. I'm in the full-stack program and this definitely motivated me to sign-up and motivates me to finish it, too.

Thanks. Did not realise this. Though $500 is still a huge chunck of change for people here. Considering the exchange rate

What I find ridiculous in this article is how bad and social focused Google hiring has become. Google regularly declines to hire people with pHds or decades of real dev experience for developer jobs, yet hires a person after one "full stack developer course?" Reading between the lines we can see how this really works: female: +1, ivy school: +1, young: +1, etc.

On the topic of udacity, I knew from the beginning these claims that they were going to replace traditional education were equally ridiculous. Their courses might get some people jobs they can't handle, and those organizations will subsequently fail at their attempts to develop products. I see stupid stuff all day long. My previous employer hired (IMO) unqualified engineers for (IMO) social reasons. Those people did not commit anything resembling code for months, but of course social politics of the workplace prevented them from being fired or anyone saying anything about their non-work.

Traditional education is traditional. This means everything else has been tried. Everything. MOOC's are not dissimilar to paper based correspondence work that has been around somewhere on the planet for probably hundreds of years. When it gets down to reality, the surviving companies are not going to say, "you watches some videos for a few months, welcome aboard." Those companies that do will slowly die off.

The "full-stack developer" nanodegree mentioned in the article appears on Udacity's list of nanodegrees but, when you click through, it takes you to a page entitled "Full Stack Web Developer Nanodegree".

The curriculum seems decent enough[0] but it's focused on a particular skill set (building and deploying server-rendered web apps and the APIs to support front-end code). It doesn't cover front-end development (e.g. JavaScript or specific frameworks like Angular/React), and so it falls short of 'full stack web', and further short of 'full stack' in the original sense.

[0] https://www.udacity.com/course/full-stack-web-developer-nano...

The "full stack" starts with electrical signals a thousand miles away and ends with glowing pixels in front of you. Javascript is a teeny tiny sliver of the "stack".

That's what I meant, but I could have worded it more clearly:

- The lack of JS/front-end means it's not really 'full stack web development'

- The lack of lower-level stuff means it's definitely not 'full stack development'

I'm not sure, though, that you and I would draw the 'full stack development' line in the same place. For me, the lowest level would be OS kernels. For you, perhaps it would be CPU microcode or even deeper.

If someone could write an OS from scratch (running on an off-the-shelf hardware) and make that thing respond to HTTP requests, that would be sufficient to call them a 'full stack developer', right?

Well, and render what was returned from that request onto a screen :-) But a lineprinter would do, I'm old-skool like that.

Udacity (and Coursera, edX, MIT OpenCourseware) can and do teach very helpful skills to large groups of self-directed learners. I have benefited from Coursera. I suspect that most who come here fit into that group. One problem is that a large proportion of the population do not have the personality/skills/drive to do this. I would note the Udacity pilot with San Jose State University. They started with a remedial math course, a college algebra course and an introductory statistics course. The pass rate of the stats class was about 50%, the two other courses much lower - 27 and 25% if I remember correctly.

One approach that has worked better is a partnership with a Community College (e.g. Bunker Hill in Massachusetts). Here they use a flipped classroom approach. The students watch online lectures and come to a lab session where they have an instructor to help when they get stuck and are encouraged work in groups. This seems to help many get past the hard parts and complete the work. This approach is obviously more expensive. Very few people work "for free." Self-motivated, disciplined individuals will always do well with systems that focus heavily on independent study. Others will have to pay more to access more personalized support.

Reads like a PR written article. Yes online courses are real, but article almost solely focus on Udacity and its success story.

This is much broader trend (coursera, teamtreehouse etc.). There are also some obstacles (motivation, huge enrollment though very low percent of course completion).

I wish journalist do their research, instead of writing semi-ads.

Anyone can teach fast, the question is does anyone they teach learn and become productive?

The perils of 'java' schools, written ten years ago and still relevant http://www.joelonsoftware.com/articles/ThePerilsofJavaSchool...

Not everyone can be a productive programmer, independently from what they want, how hard they try or how they are teached

If a man selling guitars and guitar lessons says he has a sure fire way to make you a rockstar, check your wallet.

What did Udacity say that is comparable to "sure fire way to make you a rockstar"?

> It Can Teach Tech Skills to Millions, and Fast

I would say that too if I could charge everyone of them $200 every month.

U's cs101 was a game changer for me. Loved blazing through on demand.

U shot itself in the foot with bad PR when it publicized stats that showed lots of folks starting but not finishing classes. Haters jumped on that. .. But honestly i never saw the big deal. They're like textbooks... you get something out of it but seldom read every page. So what... you're learning!

I read this PR piece as U having learned from that horrible move a few years ago.

Motivation not included.

This is great news for anyone who want to learn something to create something but it will hardly help anyone who want to get a job.

Kind of ironic isn't it? If you can create something of value from scratch but still can't get hired what are companies hiring for?

Problem is that what you can create from scratch isn't what the companies often need. Instead they need experience in working with something already created which is quite a different monster IMO.

You're right but one can quickly lead to the other if the company is willing to dedicate even a small amount of resources to training.

Yes and there in lies the issue. They normally aren't in my experience.

I looked at their full stack program: Apparently for my startup I've done most or all of that already! So, right, I've got up my Web site, with Web pages with HTML and CSS, and four server side servers. One of these is just SQL Server. Another is a Web page session state server I wrote. And the other two are specialized to the technical internals of my site.

I did it all from self-teaching. Gee whiz.

"Look, Ma: No courses!"

Also looked at their program for training data analysts to "Discover Insights from Data".

"Insights"? For that goal, there needs to be some caution as in, say,

     Statistical Science
     2001, Vol. 16, No. 3, 199–231
     Statistical Modeling: The Two Cultures
     Leo Breiman

as at

http://projecteuclid.org/DPubS/Repository/1.0/Disseminate?vi...

Part of their work is using exploratory data analysis. Gee, there's a book with that title by J. Tukey. I have a copy of the original, that is, the manuscript, that is, in Tukey's handwriting! It was okay.

They mention finding "anomalies" in data. Okay, how about an anomaly detector that is both multi-dimensional and distribution-free with false alarm rate adjustable in small steps over a wide range, say, one a minute, hour, day, week, month, quarter, and known quite accurately in advance? I published one of those.

Gee, guys. I also sent 1000+ resumes and got back zip, zilch, and zero.

Are some people hiring? Do they really know what the heck they are looking for? Will they recognize what they are looking for when they see it? All maybe not.

Reply on news.ycombinator.com