As easy as it is to find information online, the internet is messy, and seeking to learn a topic by googling it and reading up as much as possible can be a haphazard and frustrating process. The arrival of e-learning is seeking to change that. By having focused, passionate educators collate their material in one place, it has suddenly become much easier for any driven students to find all the material they need clearly presented in one place.
The excellent Khan Academy already exists to cover all of maths, and much science, to secondary level, and now the universities are starting to arrive in the game. MIT has had its open courseware around for a while, and Yale has a smaller selection, but neither seems to have caught on in any big way.
A major new experiment took place last term, organised by Stanford University. They announced three classes that would be running online, with the flagship one being Introduction to Artificial Intelligence, taught by Sebastian Thrun and Peter Norvig. This new course represented a big departure from the old style of simply recording lectures during the term, and dumping them online.
First, it borrowed some of the best features from Khan Academy. Hour long lectures were replaced by much shorter videos, recorded specifically for the online contingent, most lasting less than five minutes. Combined with the fact that these were interspersed with quizzes, where interactive software allowed you to check boxes, or enter numbers in answer to questions, this made the whole experience far more engaging and easy to follow.
For the first time, the class was being run exactly parallel to the actual version being taught on campus, with all the same material being covered. This meant that there was a real time limit set on the class, as opposed to being able to take some time off, and return back to the same point whenever you felt the inclination return. Enrolling in the course meant making an active commitment to set time aside each week to watch the videos and learn the material.
It also meant that there was an online community of students all taking the class at the same time, adding strongly to the sense or realism, whilst also providing a group of people to ask questions to, and to provide help, if you struggled with a particular topic. Perhaps most importantly, they acted as your competition. At the end of the course, you would receive a letter of accomplishment from the two instructors (although not endorsed by the university), which stated your rank within the class, based on a combination of your homeworks and the two exams.
In ten weeks, the course sought to provide a brief overview of the many different topics covered by AI. The first half was more theoretical, dealing with search algorithms, formal logic, and the topic I enjoyed most, probability, more specifically learning about Bayes theorem, and its uses.
After the midterm exam, we took a look at practical applications, namely game theory and planning, computer vision, natural language processing and robotics. The highlight in this section was the chance to hear Prof. Thrun speak about his own work on driverless cars, and his experience in winning the 2005 DARPA Grand Challenge. Watching videos demonstrating the impressive skill of these vehicles, while learning the principles that governed them felt almost futuristic in a sense, but also represented what could be the next area where AI research could make big impact. Google has already used driverless cars in creating its Streetview, and it seems likely that this technology could start to diffuse into the mainstream very soon.
What then were the results of this ambitious experiment? From a personal standpoint, I am happy to call it a promising success. This really was a first of its kind, but managed to be hugely enjoyable and rewarding. Credit for this must go to the two professors, whose enthusiasm for their subject translated clearly into their videos, especially during the virtual office hours, when they would take the time to answer submitted questions, and where their broad knowledge and expertise became apparent.
Obviously and understandably, there are many improvements to be made. The setup wasn’t yet there to incorporate programming assignments, the only difference between the online and campus versions. Some of the earlier videos weren’t particularly smooth, and it can be hard finding good questions that can only be answered by multiple choice or numerically. The site also had a habit of crashing near homework deadlines, showing that some things, like last minute scrambles to finish work, will always stay the same, whatever the medium. The final exam (pdf) was disappointingly easy, but if that’s the kind of thing I’m complaining about, then I certainly shouldn’t be too upset.
More than 20,000 people completed the course, but most of the statistics on the class are still unkown. Prof. Norvig is set to speak at TED in February, where he will hopefully provide more information.
This project has hopefully cleared the way for this type of experience to become more frequent. Stanford University are continuing with a new set of classes (do sign up, you have nothing to lose) available next term, and it will be interesting to see how other universities will respond.
It has also prompted a lot of questions to be asked about the future of education in general, and of the role of universities in particular. Is the best use of a teacher really to stand in front of a class and deliver lectures with minimal student interaction? Will it become possible to receive more official accreditation for completing this type of class? Will there still be the same demand to attend prestigious, expensive universities and schools if all the material is online? How many professors will have the enthusiasm to run an online course successfully?
However these questions eventually get resolved, I am confident that an improved education structure will emerge, where high quality content is more accessible, and where it is far easier, for far more people, to achieve their potential.
No comments:
Post a Comment