Thrun pivots Udacity toward vocational education | Hacker News Reader

Thrun pivots Udacity toward vocational education

cis471.blogspot.com

118 pointsAmadou95 comments

The key quote from this article is right here:

"I think Thrun's elite background led him down a garden path. Any San Jose State professor who had taught an introduction to statistics could have told him that many (most?) of his students would not have basic arithmetic skills and would "hate math." They are not Stanford students."

Thrun is an elite academic. He has most likely never even attended a university that had to teach remedial math courses, let alone taken or taught one. Thus, by extension, he has likely never had to deal with students who had anything but excellent preparation for studying the material he was teaching.

I sat through most of Thrun's AI lectures and I felt that he and Norvig were questionable teachers. They are clearly brilliant men, and their excitement was contagious. They would be tremendous guest-lecturers in any CS course (the particle filter lectures were Thrun's best because he was so excited, but he spent most of his time gushing about how cool the ideas were). But as far as teaching a course to non-Stanford students without the help of an army of TAs, they were uninspired.

Compare Thrun's experience with the sentiments in this article about Miami University in Ohio: http://www.propublica.org/article/on-country-club-campuses-a...

The common vein is that education is fairly easy when the people you are educating are well-prepared and generally hail from the upper half of the socioeconomic spectrum.

Teaching poor kids, and kids who, for other reasons do not have adequate preparation is really something of a Sisyphean task. The problem is not in our schools, and it can't be fixed by fiddling with the curriculum or delivery method, it can only be truly fixed by fixing the underlying social problems, but that usually means talking about sticky issues like racism, sexism, ethno-centrism, and capitalism itself.

I'm sure I'll be downmodded for pointing this out, but your post also suggests why these disparities persist even after college.

Via a sisyphean effort, colleges manage to graduate a group of unintelligent and unmotivated students. But after graduation those students are still unintelligent and unmotivated, they simply had knowledge of calculus forced into their heads for long enough to pass an exam.

As an employer, I need employees who will show up and start building shit. Walmart and McDonald's have time for heroic efforts in extracting value from employees, I don't. This is part of the reason why github is such a great hiring tool - pushing good code all by yourself is a strong indicator that you are intelligent and motivated. Graduating college, not such a good indicator.

Interestingly, if colleges didn't engage in heroic efforts to get their students to graduate, college would be a much better hiring filter.

I personally find this comment to be incredibly condescending. It smacks of misplaced elitism.

It comes across as though you are a bad manager who simply doesn't know how to nurture your new employees and you are trying to blame them.

If you don't want to nurture young people straight out of college, most of whom have simply been told exactly what to do at each stage, don't hire them. Hire experienced professionals.

That's why they cost more.

I'm not trying to blame anyone and I have nothing against inexperienced people. I'm quite happy with the people I've managed in the past, most of whom have been fairly inexperienced. They were generally better employees than I was a manager [1].

There is a difference between nurturing and handholding. Nurturing is opening the door for them, handholding is pushing them through it.

A concrete example. I had a young college kid put in charge of a bunch of mechanical turks, most of whom were considerably older and more experienced than her. Her first few days of the job were rough - the older girls thought they deserved the job and gave her pushback and I was also unintentionally undermining her. Nurturing her involved asking what she needed, redirecting all work queries to her, and building a report to let her change the business process (she switched from time tracking to goal tracking).

Nurturing was simply telling her underlings "ask little boss lady" when they came to me and then going back to coding. That's exactly what she asked me to do. Handholding would have involved "hey everyone, I need you all to be nice to little boss lady and if you aren't I'm going to yell at you."

(Incidentally, if anyone needs a great manager in Mumbai, let me know. I think she's on the market.)

[1] I failed to protect the people below me from the people above me and I can't blame that on anyone else.

It may have sounded harsh but most employers aren't right on the bleeding edge where they have the pick of graduates with Firsts from tier one universities.

Hint Google etc probably dont get many of the top Grads these days they get the layer below who put money over cool research.

Well, in your particular case, I guess just having a lot of shit on github would be sufficient :)

This sentiment seems reasonable, but it is has pernicious roots. Thrun was "led down the garden path" - he wasted time with a frivolous idea. And he did it because he was fooled by his elite background, and by implication his inability to see how that elite background biased his thinking.

In other words his failure is due to personal failings or shortcomings.

Another interpretation, one that I prefer, is that Thrun tried something bold and novel based on a perfectly reasonable hypothesis. It didn't work, he's learning and adapting. No garden paths or greek tragic flaws required.

I should say that I don't disagree with glesica's points about the difficulty of educating people who haven't been groomed for college since birth. I just think there's far too much unwarranted and barely-disguised smugness in much of the online commentary.

>Another interpretation, one that I prefer, is that Thrun tried something bold and novel based on a perfectly reasonable hypothesis. It didn't work, he's learning and adapting. No garden paths or greek tragic flaws required.

The problem with this is that there are real (underprivileged) human beings used as guinea pigs for this "reasonable hypothesis".

Thrun can fail all he wants on his own dime and time.

http://tressiemc.com/2013/11/19/the-audacity-thrun-learns-a-...

The message from the link seems to be that failure is not acceptable for anybody trying to help less privileged kids. But what's the alternative? Don't try anything new?

The author, and Thrun's critics within the existing system, often have hypotheses of their own for ways to improve the situation, but don't those carry a risk of failure as well?

Say a sociologist were to provide a cadre of students with free housing and food and an allowance, to test the hypothesis that financial stress contributed to academic underperformance. And say it didn't work out: maybe those students underperformed even more, due to some confounding factor. Would Tressiemc accuse the sociologist of a war-crimes level ethical breach, as he has done Thrun?

"Thrun is an elite academic. He has most likely never even attended a university that had to teach remedial math courses..."

Thrun went to college at University of Bonn. I don't know how it ranked 25 years ago but today it's # 178 on Times list, around # 75 in Europe and only the 9th highest ranking in his native Germany: http://www.timeshighereducation.co.uk/world-university-ranki...

Not that I give much for these rankings but University of Bonn is hardly an elite university. It's a good public university, just like so many other good European universities. But I would be surprised if not some students at University of Bonn take remedial math courses.

The Max Planck Institute of Mathematics is in Bonn - I'd imagine that it's influence extends to the university in terms of top-tier mathematics education.

The difference of universities in Germany is far less than the difference of US universities (thus terms like "elite universities" are hardly used in Germany despite tries of the German Government to give some universities additional funding to become something officially called "elite universities").

There are really hardly any remedial math courses in German universities since vocational education is rather common in Germany if you don't want to make a general qualification for university entrance (Abitur). Additionally in Germany (that's very German) there are universities of applied science (Fachhochschulen) which give a very applied course of studies. There are even some more more exotic ways of getting higher education (vocational academy/university of cooperative education (Berufsakademie)). Thus in Germany you go to university since you want this kind of education, not because there are hardly any alternatives. Living and having studies in Germany I never heard of something like remedial math courses - they say if you aren't able to self-teach such material you should rather not go to a university or at least be a student in a different subject.

I'm sure that most US elite universities (maybe not Stanford) have a goodly number of undistinguished legacy students and humanities-focused undergraduates who "hate math". They're just dealt with through the aforementioned armies of TAs, the drift into easier or less mathematical majors, and of course grade inflation before they'd start to bother someone at Thrun's level. A TA or associate professor at an elite university might have a much clearer view of what undergraduates everywhere are like than Thrun.

This isn't surprising, really. MOOCs disproportionately benefit those who already have the means, the background, and most importantly the drive to learn.

The greatest challenge of education is to nurture the desire for learning in students and for students to be in a situation in life (e.g. situation at home - income, safety, etc) in order to even have a chance at becoming motivated.

MOOCs address the problem of access. But access wasn't the main problem to begin with for the vast majority of students whom people like Thurn was trying to reach.

It's worse than that: MOOCs don't increase employability even to someone who has the "means, the background, and most importantly the drive to learn."

It's like reading a book: No one else give a shit.

I would add that for most jobs knowing more than the basics from a course, and often not even that, isn't useful.

Books got me a few jobs.

Indeed. I got a job offer merely on the basis of having worked through much of K & R.

Reading books is not a reliable way to get a job, certain university degrees are, as are certain types of work experience and certain accreditations.

If you get college credit and eventually a degree then people do give a shit.

Which is the point: It's not book reading itself, it's the (valued) certification.

Reading a book won't get you a job. Being a person who reads books may very well get you one.

Not at all true. My girlfriend, a Stanford graduate in mechanical engineering, used MOOCs to do a career transition from product design to computational disease research. She would never have been hired without the skills she learned online.

I disagree. I can't quantify it, but taking a few impressive-sounding (and -being) courses makes you look good in employers' eyes. It shows you to be a self-driven learner who tangles with hard stuff. At least somewhat. But you're right in that it's soft; a hint, rather than a certification.

Wouldn't any educational effort disproportionately benefit those who have the means, the background and the drive to learn?

What you are describing is not education, it's convincing people to get education, which is completely different issue not solved by MOOCs or any other education technology.

Also, given how expensive currently education in the US is and how it still does not have any need to reduce the prices - on the contrary, educational prices are raising way ahead of inflation and most other prices - I wouldn't say at least US society has the luck of motivation to get educated as a widespread problem. Some people surely do, but how many? How many of them are similar enough to be captured by any single approach or narrow set of approaches that a single company could cover?

> Wouldn't any educational effort disproportionately benefit those who have the means, the background and the drive to learn?

No. New Zealand's Reading Recovery program is one example that springs to mind. Since it focuses on providing one-on-one tutoring to children who are severly behind in their reading (e.g. 10 year olds who read like 5 year olds), those who ebenefit defintely tend to have problems with the means, background, and drive.

Awesome example. But that's a targeted intervention. Any treatment/course that's provided universally will tend to help most those best able to use it, i.e. the best prepared, most privileged, etc.

This is part of why equality of opportunity can't lead to opportunity of outcome: Differences in ability and character exist and the better the background environment the more they matter. It's like height. When there's substantial variation in nutrition a lot of variation in height can be explained by that but once more or less everyone has adequate nutrition there's still a hell of a lot of variation in height, and it's due to genes and randomness. Environmental influence has been reduced to close to zero by equalization.

You are correct, in that if you are talking about remedial programs, they would benefit those who have already failed (by the virtue of being remedial programs) to achieve something that is considered normal to achieve. However, if you're talking at general education programs, not specifically targeted and limited as remedial ones, I think my point still stands. I'm not saying remedial education is not important, but I think it's a different field - like a gym and a hospital, both aim for people to be healthy, but they address different populations by different means.

"The greatest challenge of education is to nurture the desire for learning in students"

I could not agree more. I think the idea of a curriculum is quite broken, especially in high school and below. I'd much rather have a huge list of stuff that will be tested upon university entry and let students learn whatever they want during high school (well ideally I'd want universities to do 1 on 1s with every entrant) and spend 100% of the time on creating an environment in which learning and exploring what interest you is encouraged.

I think the fundamental problem is that most people (including myself) have some assumptions on what a good education is supposed to entail and have a really hard time not trying to "force" kids to learn certain things. I blame Plato :P

I fiercely applaud Thrun's update on the uncomfortable evidence he gathered, and wish him the best in his pivot.

(It still seems to me that MOOC's might be done right with enough effort, like spending $100K per hour of nationwide-reproduced instruction the same we do as with TV, but I have not been keeping up with the literature here and perhaps I don't know how pessimistic I should be.)

Someone with deep pockets and deep understanding would need to lead that effort.

Actually, just someone with deep pockets and deep care. That person can hire experts in both the field being taught but also in video production.

A 40 minute TV show (scripted, non-reality) easily costs $1-3M to make.

I tried to produce some online learning videos recently. Doing it myself took 10 hours of post-production work to create 1 minute of final content. After making 10 minutes of content (which took over a month), I had to stop for my own sanity. [Actually, this post reminded me to finally upload them. They're currently uploading to http://www.youtube.com/user/learndofun if you're curious.]

For some educational material (especially intro content), you want well produced, easy to follow material. The higher up you go, you can just throw a camera in a room and people will be happy with things like http://videolectures.net.

Perhaps public-minded funding would be better spent on ways to make the production of educational material more accessible and efficient.

Aren't khan academy videos pretty cheap to produce, but quite good( except for some pedagogical errors khan is doing because he isn't a teacher, but that's easily solvable) ?

> except for some pedagogical errors khan is doing

care to explain as I'm confused?

Khan's lectures are a bit extemporaneous in an attempt to be more conversational. Because of this, he often makes errors or explains topics sloppily.

It's not so egregious that it ruins the content, and it seems to fit his approach of fun, approachable lectures that will lead the student to explore the topic further with classmates, instructors, other internet resources.

Also, I wasn't to fond of using black background and bright colored squigly writing for learning calculus. I have ADHD and learning math and calculus has been such a huge challenge. But I found work done by patrickjmt[1] to be amazing. It's clear neat and concise, and was able to get through Single and Multivariable thanks to him. My girlfriend and my sister really loved the way he explained calculus as well.

http://patrickjmt.com/

There's some stuff he doesn't explain properly, and he makes some assumptions about prior knowledge which aren't really valid.

That said, most teachers will also make the same mistakes.

He's not a great teacher. He's a good teacher who uses a new medium very effectively.

He's not a great teacher.

By what measure. He's better than every teacher I had from K-12. Better than most of the Profs I had at uni as well.

What was your experience?

Given the popularity of Khan Academy, I don't think it is fair to say that Khan "isn't a "teacher" anymore. Also, yeah, they look pretty cheap to produce.

That's not analogous at all. A TV show has to pay for multiple professional actors, sets, special effects, a large filming crew, etc. Post-production/editing costs would be a relatively small portion.

Top-notch educational materials require a single instructor and a single camera, and no matter how much production you do it should still come out substantially less expensive than a scripted TV show.

>Top-notch educational materials require a single instructor and a single camera, and no matter how much production you do it should still come out substantially less expensive than a scripted TV show.

Educational content can benefit by the use of props and effects as much as any TV sitcom; and the props are much more expensive.

Not to be rude, but not only is this over produced, the level of production makes it pretty incoherent.

I get that you're trying to illustrate your point with planes matching up to your hand movements, and blocks of information entering the frame when you refer to them. It's just a confusing mess.

They're designed to be frustrating at first so you'll figure out what you're missing.

You know you know the material when you can understand everything at the speed it's presented. (Chopping everything in 2-5 minute blocks helps with that too. Worst case, memorize everything and move on.)

[also, the posted content was part of a second level linear algebra course, so a lot of it is presented as a series of bullet points with just enough detail to jog existing memories of terms/definitions/how things interact.]

Also, that's just some kind guy who has a PhD in particle physics at Princeton who volunteered to let me record him saying things with no coaching or guide at all. There was zero coordination between the hour he talked while I recorded him versus the month I spent adding incoherent flying equations everywhere.

You know how Michael Bay movies are horrible because he wants to throw in every CGI effect he can? It was like that. I wanted to experiment, pedagogically, with different kinetic visuals. I feel my knowledge spatially/visually, and video is a perfect medium for toying with spreading "how I think."

You should have seen the first version I had with tons of explosions and racially stereotyped eigenvectors.

I know linear algebra and I was still very frustrated.

Success!

There is a chance I overestimated the amount of frustration people who aren't me want to feel while edumacating themselves.

Don't worry. The project is dead. There is no fear of me trying to teach you anything that way in the foreseeable future.

For what its worth I think it's an effective method for teaching. Now of course the execution of these particular set of videos is questionable, but I think the idea behind it is a sound one. I'm also a very visual learner and seeing things play out is very helpful. The first two videos were on the right track with having dynamic equations that represented the concept being talked about. The third/forth videos didn't seem to have as meaningful of graphics/equations so the benefit wasn't as great.

Also, seeing you in the video was just a distraction. Especially the cuts to maintain verbal flow were jarring because of the obvious visual cut. Just a whiteboard of equations like kahn academy would be better. If the idea here is to overlay graphics over any lecture that would also work, but then you'd have to either switch between live lecture and whiteboard or just have a whiteboard on the side.

These are great! I think there is definitely a place for MotionTex in math videos.

Thanks! The secret sauce is http://www.chachatelier.fr/latexit/ with export to PDF.

An online course costs somewhere between $45k and $400k to produce[1]. UPenn spends about $50k on each Coursera course it puts up (Coursera courses are much more akin to simply videotaping lectures). While Udacity spends about $200k on each course (which are usually of a higher production quality and more interactive), and expects to double that for its Online Masters in CS with Georgia Tech to $400k[2]. EdX supposedly charges universities $250k for design and consulting services for putting together an online course.

[1] http://www.educationnews.org/online-schools/moocs-may-be-fre... [2] http://www.omscs.gatech.edu/

This is consistent with my limited experience in video. Tons of editing is required.

This makes the Khan Academy all that more impressive.

Isn't this just them trying to monetize online classes? I've read a few other articles that covered this topic and it seemed he was pivoting away from "education should be free" to wanting paid offerings. I know he has, along with others, complained about the limited number of people completing courses or passing them. I just feel like that's an excuse. I looked over a number over courses but didn't complete them because they were boring, lacked necessary prerequisites or had a poor user experience. I think it's a little early for giving up and I hope something takes free online education seriously enough to solve the problems. I'll also mention that not a single person I know outside of the tech industry has even heard of this stuff. You have to at least try a little before you quit, right?

When you learn for the sake of learning, rather than certification, there's little point in crossing all the t's and dotting all the i's, in entirely finishing a course. You take what you need from it, and move on. Lifelong learning precisely means that you mix education in more closely with the application of the information you have learnt. When I use these resources, I jump around from course to course because I am actively engaged in my education, trying to get exactly what I need to move on to the next step.

If MOOCs want people to follow the material point to point, from start to end, they will need to provide a level of certification which gives people a meaningful incentive to do so.

The education is still free. They're still offering all the course materials they did before, they're just adding on the mentoring and certification tier.

Did anyone take Thrun's basic stats course? Is it the same one that Udacity offered to the public? I'm bothered by this paragraph: "I think Thrun's elite background led him down a garden path. Any San Jose State professor who had taught an introduction to statistics could have told him that many (most?) of his students would not have basic arithmetic skills and would "hate math." They are not Stanford students."

San Jose State is a reputable university with many graduates doing well in the valley. The statement that "many of his students would not have basic arithmetic skills" sounds shocking.

It's "shocking" (really?) if you don't realize that those same students without basic arithmetic skills might not graduate, which is not inconsistent at all with the notion that current graduates of that university actually do very well.

Thrun is right if his point is that it would be nice to bring them up to speed in a more thorough way, rather than just weeding them out.

I did realize the survivorship bias. It's shocking to me because if a good university like SJSU would have so many incompetent students, something is wrong with the K12 education system, or something is wrong with American families.

Both.

You state that you are surprised that a school with good graduates has incompetent students, which implies that you wouldn't be surprised if a "bad school" had incompetent students.

However, that logic implies that San Jose State has a lot of selectivity in admissions, which it does not: http://en.wikipedia.org/wiki/San_Jose_State_University

It accepts over 60% of applicants. And frankly, a mark of a most useful school is one that isn't too selective in admissions and can churn out good graduates.

Not to blow your mind but: Try to challenge some of your med students (or grads) to complete it and see how they do.

I'm not sure what your expected result for this is: that they'll ace it, or do just as badly.

At SJSU, 33% of the incoming freshmen need remedial math and 49% need remedial English. It's like that in most of the Cal State schools.

http://www.calstate.edu/pa/clips2007/march/14march/unprep.sh...

at every school in the world, there are kids that need help with math. Even at MIT there will be some kids that have been memorizing formulas... that's simply because that's how they've been taught for their entire life. Then they hit a wall and they're really in trouble.

Thrun can't teach kids that have, for a decade, been learning how to memorize crap. How many math teachers in high school and middle school can prove the pythagorean theorem? That should be the survey that we worry about.

Thrun should be making courses for fifth graders also if he wants to teach 20 year olds intro to stat.

Then the education system has a serious problem. Kids respond to incentives. Memorizing formulas would work only if students can get by homework and exams with rote memorization. We probably should learn from Russia/India/Korea/China, where it's almost impossible for students to pass STEM exams with mere memorization. Not that their exams are hard. It's just that their educators take effort to make sure the problems in their exams encourage intuitive understanding, original thinking, and creative problem solving. Ironically, US universities do exactly that in their classrooms. I wonder why high schools can't just do the same.

Umm.. India? I am from India and rote memorization of math formulas is very evident. I can not speak for other countries but you can pass all the exams till high school using rote using rote memorization and get 100 % too. Only exams like JEE and to an extent AIEEE actually test your ability to apply concepts and that's why they have the smallest selectivity ratio.

Welcome to the shocking world of modern public education, where every fourth student[1] at 12th year of education can't read at a basic level.

http://nces.ed.gov/nationsreportcard/pubs/main2009/2011455.a...

I do not see this as a problem with Thrun, but as a problem with the current educational system. If MOOC's carried a fraction of the "academic credibility" that college degrees did, Udacity would be a home run already. This will require a cultural shift, and companies like Udacity are sowing the seeds.

As Warren Buffet has said, the courses he took at Wharton barely differed in quality from the ones he took at Nebraska. Our culture is so obsessed with shallow accolades that it is slow to recognize a humongous opportunity right underneath its very nose.

There is absolutely no reason that MOOC's cannot teach many college-level courses as good as or better than their real life counterparts. Unfortunately, the motivation of students largely stems from the aforementioned shallow accolades.

As a filtering mechanism for professional skills, I think the current academic system has some flaws which MOOC addresses nicely, perhaps not as a complete replacement, but a powerful complement. But the reality is, academia should be used to filter academics, and in many industries, a degree represents nothing more than 4 years that could have been spent learning the actual profession and adding value to the economy.

> If MOOC's carried a fraction of the "academic credibility" that college degrees did, Udacity would be a home run already.

It is unclear to me that any educational system that passes such a small percentage of students in Elem Stats would ever have a hope of gaining academic credibility. That's the playing field.

What an instructor at SJSU would do that Thrun is not going to do is lots of office hours, question-answering, and mentoring/being there. That's what it seems to require.

Personally I think it took a great deal of courage for Thrun to recognize that this is a game he is unlikely to win.

>It is unclear to me that any educational system that passes such a small percentage of students in Elem Stats would ever have a hope of gaining academic credibility.

Do you think that if they inflated their pass-rates that you'd find them credible?

There is too much focus on pass-rates[1]. I shouldn't have to point out to a person who implies having taken statistics that it is silly to compare selective institutions' pass-rates to the pass-rates of open-enrollment institutions. Add the benefits of "free" and "I don't have to leave my home/workplace/etc." as is the case with MOOCs, and you have the potential for greatly expanded enrollment. Why would you think that large enrollment growth would generate better pass rates?

[1] I'm an instructor at a two-year college / vocational college where some people seem to never shut up about pass rates. The State also seems to think that pass-rates == quality, and has set about incentive-izing programs with "good" pass-rates. Even worse, one local (large) employer insists upon "B" or above average for anyone it hires.

>Personally I think it took a great deal of courage for Thrun to recognize that this is a game he is unlikely to win.

Maybe he's not in it for the money, but to make genuine improvement.

> I'm an instructor at a two-year college / vocational college where some people seem to never shut up about pass rates.

Yes, those people comprise the playing field.

The article makes it sound like he gave up real quick. That he wrote class materials at a level above where the actual students were and instead of adjusting to their level for the next semester, he just threw up his hands in frustration.

Is there more to the story?

Sure, the article doesn't mention it but there has been other revelations of MOOCs as not being a silver bullet or what they were hyped to be, not just Udacity.

Yes. Read the original article (http://www.fastcompany.com/3021473/udacity-sebastian-thrun-u...) this is based on. It's much more informative.

Thanks, I saw the link to it and didn't realize it was current, I thought it was just background on Udacity before this pivot. I probably should have submitted that link instead.

As an academic I feel no schadenfreude at all. I feel somewhat sad.

This model will work eventually. It's mostly a matter of how the courses are structured and taught. I don't think it'll replace traditional universities but it should challenge the price structures and general lazy-bureaucracy attitude of many of them.

If you look at programming in particular I feel very good about all these online courses. It's easier than ever to teach yourself programming which is a net+ for society. I don't understand how more accessible education for everyone is ever a bad thing.

I don't think it's just education that's broken though. HR is also broken because they mostly look for the signature on the dotted line (oh look someone else has done the vetting for us) and tend to value certification over actual skill. Long term, I hope it'll be enough to list coursera etc. courses, books you read and a github link on your resume. IT should strive to be a leader in this shift because it's generally an industry where self-learners can do extremly well and have traditionally shown it's possible.

>Long term, I hope it'll be enough to list coursera etc. courses, books you read and a github link on your resume.

It sounds nice but problem with this is that you get so many people who join courses just to show off certificates and end up learning almost nothing. If you include books too, there will be popular posts like "books you should read and have on your CV!". This totally spoils the idea of reading and learning for its own sake which is what education should be all about.

Well, all this hype was so overblown and the anti-university jibber jabber has been at fever pitch for some time now. It's nice to see practical thinking rather than breathless "down with universities because Internet" pronouncements.

CEO wants business to survive. They didn't get it right, ran out of time. The potential is still there, it just won't be this model.

Because the university approach is working perfectly. No humongous debt or diminishing returns issues at all.

..The current university system is just as untenable. It's just established.

The "secret sauce" of the MOOC is the on-line community of motivated learners. Can they make this work in industrial training...?

It's a shame Udacity hasn't worked as well as Thrun hoped. I'm convinced that the failure of existing online courses has been primarily a problem of motivation, rather than an inherent pedagogical problem in online coursework. It's very hard to motivate yourself to consistently work on coursework if there are no deadlines and dropping the course just means you keep going with it sometime later. Still, I can't really explain why the SJSU courses failed as well; it sounds like students had actual deadlines, and were "required" to complete the courses to the same degree they're required to complete their regular courses. It's possible that the online tools to ask for help didn't work as well as in-person interaction, or that sitting in front of a computer doing things at random times is less motivating than actually attending lectures in person at regular times. There's one thing I'm certain of, however: it's a huge waste to have universities pay professors (who are selected for their research ability, not teaching skills) to repeatedly lecture the same material to small groups of people. A hybrid model might work best. Lectures could still be online (but possibly at fixed times to prevent procrastination), but they would be supplemented by in-person recitations, where students would meet in small groups with instructors and have the opportunity to ask questions or review particular topics, with the instructor as well as with peers. This model is still far more expensive and less accessible than online courses hoped to be, but it might actually work.

I took the web dev course by the founder of Hipmunk and it was absolutely great!

This was going from "best thing ever" to "lousy idea" in light speed. That is the downside of big ideas if they haven't been validated early. Not only investment of time and resource, but psychological energy. This is really the story of all change. It it unlikely, extremely hard and often at the wrong time.

This is an interesting shift, but feels less promising than the initial rhetoric. Vocational education in STEM fields delivered by for-profit institutions is already pretty well-trodden ground. ITT Tech, for example, is a large incumbent in this sector. Why should someone go with Udacity instead?

Price, access. But mostly price. One could start Udacity while in middle school and be bad ass before leaving high school. I doubt ITT Tech offers that. Highly motivated students will excel and be able to test out of classes they don't need to waste their time with.

How is this an inverted classroom if the MOOCs don't have a component where students come in after each class and take a test showing they've mastered the lesson's material? If they don't pass then they have to get remedial correction 1 on 1. Just like in a regular class.

That's what motivates them and pushes them to work. After all habits are hard to form and break. Wanting to learn is not enough, putting in a regular effort to do so is a habit that needs to be coached. That's one of the main jobs of educators for under-motivated students (read: most no grad students). And sorely missing from Udacity.

CONCLUSION: Udacity's "experiment" blatantly omitted a crucial hidden variable.

Pity. The only MOOC I have ever finished, and truly enjoyed, was Thrun's "Artificial Intelligence for Robotics". It was magnificent to get an easy to follow introduction into a deep academic field by someone who is so demonstrably accomplished.

I know that I'm just being selfish here, but I was hoping that Udacity will continue on that path. I would love to learn celestial mechanics from Belbruno, or neural networks from Hinton, or operation system design from Linus in the accesible Udacity style. It seems that won't happen any time soon, or ever.

fyi, Hinton did a class on NN on coursera, I didn't finish it but the beginning of it was _awesome_

https://www.coursera.org/course/neuralnets

Why is it so difficult to have the motivation to finish MOOC? That's the main problem with online education, and we need to figure out what is causing so many people to quit halfway.

Another thing may be that for many people half-way is good enough. I took the Udacity cryptography course and really enjoyed it. I never took the exam because I didn't see the point. Presumably I'd have been marked down as "failed." But I think that's only because their model for learning still includes "we're the ones who certify you." I'd like to see a model that says "here's the knowledge - go as far as you want."

I would be very curious to see numbers for udacity competitors.

I know I gave up every time I tried to do something on udacity while I completed most of the classes I took on coursera, and all the original stanford ones.

I blame this on the interaction between class style and myself being terrible at self management without deadlines.

I mean, I took the original Thrun AI class with a coursera-like model and I did it fine, while I don't know if I got to the 3rd lecture of Thrun's Stats 101.

There's an interactive chart here:

http://www.katyjordan.com/MOOCproject.html

The problem with this pivot that proper vocational education also requires a fair amount of mathematics and even more tricky for an online course a lot of hands on work.

I started on the professional apprentice track back in the day and our college had two entire blocks filed with machine tools and labs.

I really wish Udacity would clarify their approach going forward. Will their classes become solely vocational, or will they continue to develop conventional courses? It's not either/or.

Thrun should look at the Khan academy.

That said, Udacity can serve a different purpose. And after all, the computer courses there are vocational...

Par for the course. Pioneers, as they say, are the ones with arrows on their backs. The "I told you so" crowd are cowards who would never stick their necks out to try anything new. That, or they have ulterior motives such as job security or income protection.

The truth is online education is here to stay and it will get better with time. It will, eventually, replace big chunks out of traditional education. Probably not all.

My son is currently going through MIT's 6.00.1x Intro to Computer Science and Programming on edX. He is 14 and in High School. Watching him progress from mechanically typing conditional statements to having his mind opened to computational thinking has been an amazing experience. Yes, this course is pushing him around and challenging him in big ways. It isn't easy.

From what I've seen, there are only two ways one could succeed with these kinds of courses: self motivation or external support (or both). In our case he has all of my support. I am actually taking the class alongside him so I can see what and how they are teaching in order to help him out.

Motivation is a huge factor. He is a member of the local FRC robotics team and was involved in FLL before that.

We work on every problem set in a collaborative manner, with me guiding rather than providing solutions as well as simply being there to expand on topics that are not covered to a great depth (pointers comes to mind). Lately I've been doing a lot more watching than guiding as he has definitely begun to think like a programmer and is solving most every problem without external help.

For me it's been an interesting review of topics I have not touched in years. Recursion, for example, is something I haven't touched in quite some time as I have not run into problems and systems that could justify the resources required when using these techniques. Playing with recursion in an academic setting and helping my son learn the concept was lots of fun.

I can absolutely see that a course such as 6.00.1x would be impossible to complete for a kid without the support of a parent. Not sure if that parent has to have domain knowledge or not. I can't be a good judge this because I obviously do and all of our conversations have taken advantage of this.

I can also see the difficulties in entering into some of these courses without the necessary preparation. Students who went through school by mechanically doing math without really understanding math tend to not do well on higher level courses regardless of whether these courses are online or in person.

There's also the case of the working engineer who might need to brush-up on skills before attempting a class. Using myself as an example, I have not used statistics in any formal way in a long time. If I wanted to take an online ML class I'd have to spend an amount of time reviewing statistics and probably a couple of other areas in Calculus.

In this sense this is where, perhaps, MOOC's do it wrong. Conventional live courses go through a qualification phase in order to ensure that the "herd" has reasonably uniform and adequate capabilities. The beauty of MOOC's is that anyone can jump in. And that's absolutely fantastic. What might be lacking is a departure from a linear model of teaching. Why can't I enroll in that ML class and, when and as required, take off in a branch and review statistics to then come back and "merge" into the main thread of the class. Perhaps this non-linear approach is what is missing.

All MOOC's are pretty much online versions of some kind of a traditional live class. Lectures, problems, homework, tests. All presented in a linear timescale and on a similarly linear schedule. A learning system that is truly after the acquisition of knowledge must work differently. It must take a highly interactive approach in which the teaching system is flexible enough to, effectively, deliver curriculum that is customized to the needs of each and every person.

This is a challenge. We have to be glad there are people like Thurn who are willing to stick their necks out, try, fail and try again. The critics are usually people who will never compromise their station in life to try and drive progress. They don't want arrows in their backs. Far easier to shoot them at pioneers, eventually you hit one or two of them and for a brief moment in time you might actually sound like you know what you are talking about. Reality, however, is quite different.

What I would tell Thurn is: Don't give up. Don't exit the segment. Try to figure out how to change the approach and make it work.

Reply on news.ycombinator.com