> I'd say this was mostly true 20 years ago, sort of true 10
  > years ago, and utterly false today.

Remarkably, I've yet to meet one of these effortlessly self-educated men of the Internet age. Checking out a couple of Calculus videos on KA doesn't make you an engineer. In fact, despite current advances (Udacity, Coursera, what have you), I'd argue that the kind of person who manages to self-educate to a sufficient degree today would have managed just as well 20 years ago.

Libraries and textbooks did exist back then, you know. Your comment comes significantly too early.

> Remarkably, I've yet to meet one of these effortlessly self-educated men of the Internet age.

In my experience, pretty much anyone who can program well is self-educated. They might have also done a CS degree, but most of them were programming on their own beforehand.

The mechanics of programming are usually self-taught, but able to write code that compiles != educated. In my experience those with degrees in mathematics or physics tend to be much better at producing value through programming than those without.

Absence of proof is not proof of absence. What has changed in the last 20 years is Internet. My son surprize me with the thing he learned from the internet.

Though it is also true that the things he learned are limited to what catched his curiosity. With classical education you get a more wider and uniform education, but also more shallow than if you dig the knowledge by yourself.

Note also that education plays only a significant role in hiring the next few years after it. After that it's what you have achieved and your expertize domain that become the most relevant.