In Neal Stephenson's "The Diamond Age", a plot device (not central), is The Young Lady's Illustrated Primer. It is an adaptive AI tutor. To realize TYLIP, hard AI problems will need to be solved. Yet, it is possible the iPad is a big step towards a simpler Primer.
"TYLIP is...a book that is powered by a computer so advanced it’s almost magical, and it teaches children everything. It does this through a fully interactive story. It teaches you how to read, how to do maths, it teaches you morals, ethics, even self-defence. ‘Diamond Age’ is a very entertaining read, mainly because of the TYLIP."
http://en.wikipedia.org/wiki/The_Diamond_Age
http://mssv.net/2006/05/01/the-young-ladys-illustrated-prime...
If you check out the video from that article (http://www.youtube.com/watch?v=hw5k98GV7po), you'll see a lot of the strides we're making in the areas of individual assessment (mathematics only at the moment), guidance (full breakdown of hints for every problem we dynamically generate), alternate explanations where comprehension lacks (list of all relevant videos next to every problem the student encounters), encouragement (badges and other game mechanics), and more.
As you said: it's ambitious, and we're far from the end. But we're making progress every day and seeing extremely positive reactions in all the teachers and students from our pilot programs.
What few people seem to get is that we don't need to fix education, we need to fix learning. And for that, we need exercises (as any mathematician would tell you; also see deliberate practice). It turns out digital exercises afford a range of interesting opportunities (both the video and article highlight several) for making learning more effective.
Interfaces can change relatively easily so there'll be a bunch of experiments. The exercise model is harder. And more interesting. (I've been trying to figure that out for a while now and discovered a bunch of local minima.) I'm very curious to see how this exercise model works out -- it seems promising from what I can tell.
Regardless of how this works out (not to say I think it won't), this development will raise the bar and the expectations -- both of which has been too low for too long. That is incredibly valuable.
Edit: I just watched the video and I am seriously impressed. This is the right direction. I will be trying this out soon.
Shouldn't you be used to not having a curriculum (and indeed be thriving) if you're really an autodidact?
The structure seems obvious to me. Videos are put in sequences by subject, with pre-requisites coming first. In his physics and finance sequences he frequently calls out a potential video you should watch first to better understand the current concept, otherwise if you know it then it will just trigger the memory anyway. My value from Khan Academy has been from looking at the videos, seeing which ones I already know, and watching the ones I don't know or need review on. I don't need any curriculum; my curriculum is whatever I want or need to learn at the time.
Maybe you want to call the organization of categories unorganized, but if you watch the videos in order of the mega-sequences it's pretty straightforward that it's a similar path you'd find in a textbook. (You don't read chapters of things you already know do you?)
I'll agree with you that Clippy is the future for self-learning. First we need general AI though. (Although Google's simple search is pretty good right now.)
> Clippy
Ha ha! Yes. But I don't think we need defer this until we have strong, general AI. There are steps that can be taken now, with more coming in the near future. None of these will be "the" solution, but they'll be way ahead of where we are now.
Example: currently schools have "remedial" programs in various subjects. A student having trouble with math is assigned to the remedial math program, but not other remedial programs. Ok, now extend that idea by making it more fine-grained, both by topic and proficiency. Algebra would have several areas, each with separate proficiency levels. Design curricula for each area/proficiency, and a reasonable assessment method for placement. Collect feedback on time taken and right/wrong to improve curricula and assessment. No AI, but way better than what we have. It's not at all practical in classrooms, but it's very practical on computers if it's done on a large enough scale to warrant the cost of producing such a system.
And thanks to you and others, I will probably revisit Khan Academy soon.
If we did away with state-controlled curricula I think that'd help in itself. Even the schools where you supposedly "learn what you want" have workbooks you have to follow very orderly to "pass the material". The nature of state-curricula leaves it up to the arbitrary decisions of boards and a handful of teachers instead of experimenting and doing radical changes and letting individual teachers do their own thing to see how it works out. But this is all for a classroom setting.
There are lots of places we can use HI (human) more effectively... The whole Stack Overflow / Yahoo Answers model is great too, adding an interface to auto-connect people via webcam/mic/screensharing would be nifty instead of a series of text messages.
As far as I could see each set of questions allows you to watch associated videos and also links to the prerequisites.
I'll go and investigate :) brb
Education's past is a series of textbooks (plus supplemental info and exercises), providing a clear path to learning. Khan Academy, such as it now stands, is a video analog of textbook chapters as discrete units all in an unorganized pile. Until more is done in the neglected areas, Khan Academy will remain an extremely valuable resource, but not a curriculum. Even if/when these missing bits are done this will only be the future of delivery of fairly traditional materials.
The real future of education is a computerized personal tutor that provides individual assessment, guidance, alternate explanations where comprehension lacks, encouragement to pursue natural ability and enthusiasm, etc. That's pretty ambitious, but not at all inconceivable. We're close enough to being able to achieve it that we should hold up this ideal goal so we know the right direction as we build the pieces.