Peter Norvig's TED talk reflecting on the online AI course | Hacker News Reader

Peter Norvig's TED talk reflecting on the online AI course

fm.schmoller.net

97 pointsColinWright15 comments

It's interesting that he discusses due dates as a motivator and as a way keeping everyone synced up. This is something that Udacity later dropped. I wonder why they dropped them and whether the change is having the effect they want.

I would be interested to know their reasons behind this as well.

Anecdotally, it was a bad design choice for me. I took a Udacity class in the first round after their launch and followed along with the due dates pretty well. The next round when due dates were dropped, I basically dropped as well. I found myself saying "I'll catch up on this week's unit next week" until after a few weeks I was too out of sync with the course to feel compelled to keep going.

On the flip side, I felt the one week/unit deadline was tough to work with at times given work and life, etc and wished it was two weeks/unit, or perhaps Udacity to even give you an option to choose a schedule: allow a) one week/unit, b) two weeks/unit, or c) whatever, etc. This way the course still has some structure and fixed deadlines, but it's a little more tailored to your lifestyle.

Choosing your own schedule seems like a great idea. I got wrapped up in more important engagements in the middle of a Udacity course and by the time I was able to return to working on it, the final exam had already been given, which kind of made me lose interest. Thrun has said in the past that AI will customize the classes the individual needs of the students, so I expect this will improve as the organization matures.

It would be interesting if they published stats on completions for each model. I know that in the end I stopped doing the pgm-class work after about week 7 precisely because there were due dates and I had conflicting priorities. As a point of comparison I started the ML-class long after it had expired worked my way through 50% of the material, did all the assignments and learned enough to solve the problems I needed. Did some cool projects for a few months and realized that I needed the other 50% to solve some new problem that I had, went back and did that.

One the reason we have the idea of a novel, text book, etc is because of restriction on printing. The reason we have the idea of an 'album' in music is again because of limitation on that media (it's not cost effective to produce a bunch of singles and you can't fit much more than an hour of music on a single album due to constraints of traditional media). Likewise the idea of a semester/quarter long course are essentially based on the limitations of various resources.

Ideally we'll see the length and structure of courses change with time. Why have the idea of a course at all? Why not just a directed graph of prerequisite lectures? (Khan does a decent job of this)

We're starting to see it form organically, but I think the best situation would be goal directed learning where you pick out a point on the graph and say "I want to be here, doing this" and then you fill in all the prereq nodes until you're there. This also allows us to think about knowledge and learning in much more creative ways, rather than choosing to major in art or computer science, you could say I want to use genetic algorithms to create generative art, now how do I get there?

I agree with you on this. I've noted previously how similar these (current) online class models are to the early movies, in which they simply filmed plays. It took a while to realize that some of the constraints that had been baked into the older format no longer applied.

My biggest pet peave is the whole notion of a monolithic college "degree". Being forced to sell your car to pay $5000 (you do the math) for a class in interpreting Winnie-the-Pooh from a marxist-feminist perspective in order to get an engineering degree is maddening.

If employers began accepting a la carte credentials, classes would move toward their real market values in a system of open competition. Monopolistic "tying" (if you want this important class, you'll have to also pay for this useless class) would break down and people could be much more strategic and fine-grained about what courses they took and what they were willing to pay for them.

And I would have to believe that most employers would rather have a candidate who had taken the $5000 the university had earmarked for the support of their staff marxists and applied it to about three more engineering classes, one tech writing class, and an accounting class.

I can't speak for why they dropped them, but I can tell you from my own experience due dates were more of a demotivating factor for me. I signed up for the AI class. I diligently went through the classes (and was learning a lot!) however I then had to travel to Japan for work. I went a few weeks where I literally didn't have a minute to spend on the classwork, so I fell behind. Once I missed a due date I lost motivation to continue. :\

But do you think you did so well early in the term as a result of the deadlines? Perhaps it had no impact, but I'd argue that a lot of people need that requirement to stay on track.

No, the deadline really had no bering on motivating me to start. Learning the subject was motivating enough for me. But once I realized I was falling behind, my state of mind changed.

I found that it worked both ways. When I started, the deadlines kept me motivated. But when real life intruded, and I had to take a break for several weeks, that fact that I was behind was a demotivator. A good middle ground might be "unlocking" the classes - there is a two week deadline per lesson, but it only starts when I begin the lesson.

Do you feel the parts you did interact with were helpful?

Perhaps they didn't do a good-enough job of selling the benefits of the course to you? And so you weren't dedicated enough. Is this possible?

> Perhaps they didn't do a good-enough job of selling the benefits of the course to you? And so you weren't dedicated enough. Is this possible?

Not for me. In fact, the classes I took, I would have never taken years ago (free or not). I didn't have the professional context nor the organic curiosity to see the value. I have this context now; this is what motivated me to enroll and ultimately, what motivates me to continue even though making time for the lessons is tough.

They made them less strict, I don't know if this equates to dropping them. It still has the concept of weeks with a course having a clear start and end.

I know my success rate with udacity over something like khan academy definitly comes from having a structure that I can work on every week.

The greatest thing about all the online courses is the pragmatism that comes with so many forms of education being presented in such a public light. We will learn so much more about teaching through these systems than ever before.

I've taken several online courses now and one of the greatest things I've learned is how I learn.

I signed up for this online course and it was great! I unfortunately didn't make it through the entire class -- work deadline hit hard during mid-terms. But, I would definitely love to sign-up and finish the course if and when it is offered again. Even now I still find myself going over new chapters in the text on my own -- good stuff.

This has been my experience as well. I signed up for both the ML and the AI class, they were very well done - but I couldn't make time every week to see it through to the end. I think in these cases a more flexible schedule would work much better.

Reply on news.ycombinator.com