That being said, if you're not in a leadership position that kind of decision making isn't what you're doing: you're most likely just optimizing on your own little anthill, so a "profit-centered" education like the article is against might make sense for the worker bees.
Another thing that I think is usually missing from the "train your workers" debate is how much variance there is in productivity, and how actual productivity is usually unknowable unless you have 2-3 months of project data for a given worker (these are assumptions). So hire unskilled contract workers, fire 80% of them, and then train the remaining 20%. They won't have the credentials to work elsewhere, and you've been able to identify the true all-stars using data on their actual work product.