For most jobs, there are plenty of people who can capably perform it in the world. 99% of the world fall into this category. Looking at outliers such as Zuckerberg and Gates is idiotic. Even though individual hiring decisions often don't go to the person who might be best qualified, it often goes to the person who is qualified enough - which they judge through an arbitrary bar such as a college degree. Blunt as though this method might be, it still is quite efficient consideirng the computational complexity of the problem. Anecdotal example: I work at an ostensibly prestigious consulting firm, and the work I do is simple and generic enough that almost anyone could do it. The hiring policy however revolves around best talent from best schools. Is it needed? No. It probably works well as a marketing tool for the firm's services, but even so it's not a great system. Question to ask is if this was an optimisation problem, would it be better to create a system with massive overhead that matches people to jobs with 100% accuracy, or one with 95% accuracy and some wastage?

I call selection bias. You meet the best of autodidacts and all of college graduates. Try balancing the field and then looking. College might be dumb, but not as dumb as the alternative.