I’m now on my third online Coursera course
Cryptography, Game Theory, and now Machine Learning
When I was talking with Scott Martin about Coursera, he shared his experience. Paraphrased: “I actually felt like I was learning a lot. Sure, it was easier than an equivalent course here at Carnegie Mellon, but that’s also partially because they do a much better job of explaining things in terms that everyone can understand”
Because they don’t require or assume that everyone taking Machine Learning has taken years of advanced Calculus and Probability math classes, they’re force to explain everything in a straightforward manner.
That’s it. Straightforward. Instead of confusing you with all sorts of unnecessary terms, they just tell you what it does.
Oh, sure, all the Greek symbols mean something. But nobody learned how to programming by writing their own operating system from scratch.
No, it’s a step-by-step process. You learn the basics, then you expand on them, learn more advanced concepts. You start by making a dog dance, then a simple video game, and you slowly build up to making something people would play. You don’t dive straight into every little detail of one little item – otherwise, you’ll never actually create something useful.
Maybe it’s just Carnegie Mellon, but that’s something that classes here seem to struggle with. What if I want to learn how to do computer vision (say, tracking an object with a webcamera) without bothering with months of extra “theories” and “theorems”.
It’s true, you need to know all of those theorems if you want to push the field forward and develop new algorithms.
But new algorithms don’t do the world any good. It’s the applications of those algorithms that bring us Google, Netflix, smartphones, etc. The fancier algorithms make them faster, but the more you’re taught about all of the fine-grained details, the less you’re able to apply it practically.
The real-world applicability aside, academia seems to have a grave problem: it’s ok for teachers to make things complicated. The topics are complicated, after all! WRONG. I could explain machine learning to an 8 year old. The fact that most college professors can’t is a damn good indicator that they don’t know how to focus on the most important things. This not only slows the education process and forces people away from thinking about practical applications – it also scares people away from entering the field in the first place. That is why we don’t have enough engineers.