It is common to think about technology as a fast-paced field. With new products, companies, frameworks and buzzwords, it sounds daunting to catch up. But as Dipanjan Sarkar suggests, not all motion is linear and forthright. He is Data Scientist at Intel and Editor at Towards Data Science. Instead, it might be more appropriate to see ourselves as fish in a school, circling one another. Not in vain, mind you. Instead, while each of us moves fast, the collective advances steady. That is how we break new ground in a sustainable way.
So as technologies flash upon our eyes, firm ideas take some time to stick. Sarkar chooses what he considers acceptable standards in next-generation algorithms. Those tried more than once until they show clear ways to add value. Those who made the successful jump from academia to the mainstream. Can they do so in Moodle? Let’s take a look.
Recommendation Engine. In a way, being on top of the industry’s latest development is to surf its learning curve. When a company releases a small but noticeable feature, we can gain some small insight. Which will inform the next one, and so on. We’ve had recommender systems for years, but the insight behind them is only popularizing now. It’s the nature of the pioneer’s trade. Tech only remains cutting edge for so long. Today we see how there really isn’t that much to them. It’s mostly about throwing loads of user data to the blender to pour out a guess of what they might be into next. Right?
Popularity Meta-Rankings. Alright, there are still a few issues to polish. An automated recommendation is a sound, if basic first step. But there are myriad possibilities machines can take on! They can even help us figure out what they are. Based on a given definition of popularity, a neural network can find success factors. Which can help determine the kind of content that is most effective in Moodle, beyond a format. Is a whimsical text more memorable than a dry video? Maybe only machines can tell.
Deep Linking. Another age old dissent that is earning general awareness involves the merits of supervised versus unsupervised learning. The former is easy to define. The latter is some kind of utopia. We can give an AI a goal, some variables and some constraints. Sure enough it will find a way. But are you confident enough about what you want? Why are you hell-bent on maximizing student GPA? If you are unable to realize there is more to life than this, let a machine let you see. The right approach can make data science go from insightful to enlightening. That is, among other things, the premise (and promise) behind deep learning.
Smart Proto-Assistants. Everyone has their version of the ideal Rosie the Robot. Some of us even from childhood. (Oh, was that only me?) The unanimous end game is reasonable. A system that helps us achieve the goal. Making things easier, preventing or diminishing obstacles and risk. We are not there yet, but AI can gain higher responsibilities over time. Today’s Moodle engine can give an assessment about the risk of dropping out. In a not so distant future it could choose the best remedial course. Then decide by itself whether to escalate the issue. And so on.
In sum, the future is, like, hard to predict. But as time moves on, the cone of possibilities shrinks down. We may not know which feature will set the tone for the future of Moodle. (Or become its deadly blow.) But following the trends and compounding on the ever desirable data science skills sound like reasonable survival strategies.