(Other posts tagged ai-course.)
I've written previously about the free "Stanford" Introduction to Artificial Intelligence course, on which I am enrolled along with over 100,000 others from nearly 200 countries.
Yesterday evening I worked through the introduction to the course.
Here is a key point report.
1. There have been the usual start-up hiccups [snapshot at 08.00 UK time 10/10/2011) about which I am not complaining.
2. The content consists of 1-6 minute "Khan-style" videos of diagrams being written and talked about by Peter Norvig or Sebastian Thrun. And there has been some clever/wise truncation in the videos so that they run at the speed of the presenter's voice rather than taking the time it took to write the diagrams.
3. The diagrams are made using mainly gel-ink pens on paper.
4. The audio is of decent quality, and both presenters speak with fluency. Neither sounds as if they are working from a script. (This Google Doc looks as if it is being collaboratively produced by users rather than teachers?)
5. I find this kind of "rough round the edges" production (complete with Thrun's dotted capital "I"s, finger-tips, and pens as pointing devices) has much more immediacy than flashily produced animations, and I commend it.
6. Most of the videos are followed by some "correct/incorrect" quiz questions which appear as an overlay to the final frame of the video to which the relate. Compare the basic terminology YouTube video with the same video and the quiz question to see how the quiz question overlay works from the user's point of view.
7. On one occasion there was some kind of server-side overload that prevented a quiz result from being submitted.
8. The quiz questions have so far offered (me....) enough stretch to be interesting, and one of them (relating to categorising the context of an AI challenge according to whether the environment is fully or partially observable, deterministic or stochastic, discrete or continuous, benign or adversarial) was well thought out. Likewise the use of a restaurant menu to illustrate one of the underlying processes of statistically based machine translation.
9. So far there has been a short (30 seconds to 1 minute) summary video to explain what the correct answers to the quiz should have been. On one occasion - a quiz relating to the translation of the English word soup into Chinese, my "things are never that simple" mind objected to Peter Norvig's correct answer. But I know what he was getting at.
10. The underlying structure is conventional, consisting of an outline, content and activities, and a summary. The is a large, growing and diffuse range of learner-driven support activities. Examples, from Rohan Aurora.
11. Logged-in participants can view a progress record showing the number of quizzes completed, and the percentage correct scores in quizzes attempted so far.
Comments