This is very true of graduate school vs not, but I'm not so sure about undergrad. A few points:
I stopped going to school after undergrad and started working, while most of my friends went and did grad school. When they graduated I was already ahead of the salaries they were being offered and had 5 years of experience. I'd also been earning twice as much as them for those five years and had considerable savings.
Also, I've hired people who never went to college. They were fairly qualified but what I found was odd gaps in their knowledge. They could write good code but then didn't know anything about say bayesian probability, so I would have to design the software for them which they could then code. They could certainly learn about bayesian probability, but the problem was they didn't even know it existed so they were pursuing a terrible solution to the problem that had a well known and simpler solution, until someone who went to college came and informed them. Perhaps the same applies to those who went to grad school, but I've never found that to be the case in practice.
"They were fairly qualified but what I found was odd gaps in their knowledge."
This very much goes both ways. A lot of recent college graduates lack practical software engineering experience, has overlooked the importance of related fields[0] and tend to be stuck in this examination mindset of individual correctness.
> They were fairly qualified but what I found was odd gaps in their knowledge.
As long as everyone knows that would be considered an odd gap in knowledge, you can prepare for that faster and cheaper than college (if not faster, it's certainly more cost-efficient). I could study endlessly worrying about knowledge gaps that would be an easy solution to a problem I currently have. There is value in splitting up the knowledge burden of solving problems. You can go wide, they can go deep.
I say this with no knowledge of what bayesian probability is because I've never studied it!
To fill in all these gaps you'd need the equivalent of a 4 year CS degree. I say this as someone who worked as a software engineer for about 5 years before going back for my degree.
You spend most of a CS degree studying the theory and techniques developed over the last 60+ years. These techniques are incredibly useful time and again.
It's the difference between spending a month trying to solve a problem or realizing the problem is actually just a version of a graph theory problem that was solved 50 years ago.
There are many positions in the field that can be filled by people without degrees. But there are also many that need someone who knows the theory and history. That's not to say that there aren't rare individuals who learn all this through self study, but they are rare.
5 years of grad school implies a PhD - in which case they should be hyper specialize in some field. Sounds like your friends just chose the wrong subject to pursue a PhD in. Normally you're supposed to end up being one of the top 20 experts in a very narrow field of study.
In generally this doesn't bear out in computer science (which I still think is barely qualified to be a department)
They are in fact top experts in their fields (1/2 were computer science and the other 1/2 were chemistry). But even being the top in your field doesn't outweigh 5 years of experience.
Maybe for their first jobs in industry. But assuming they aren't in academia, I'm betting a PhD in an in-demand CS field will be making a lot more than someone with a BS and 5 extra years experience in a few years.
Not to mention that for the type of jobs that require a PhD, the 5 years it takes is 5 years of relevant experience. And there are many jobs that are effectively closed to people without PhDs regardless of how much experience they have.
I stopped going to school after undergrad and started working, while most of my friends went and did grad school. When they graduated I was already ahead of the salaries they were being offered and had 5 years of experience. I'd also been earning twice as much as them for those five years and had considerable savings.
Also, I've hired people who never went to college. They were fairly qualified but what I found was odd gaps in their knowledge. They could write good code but then didn't know anything about say bayesian probability, so I would have to design the software for them which they could then code. They could certainly learn about bayesian probability, but the problem was they didn't even know it existed so they were pursuing a terrible solution to the problem that had a well known and simpler solution, until someone who went to college came and informed them. Perhaps the same applies to those who went to grad school, but I've never found that to be the case in practice.