My young son just started walking a few days ago. You can see from this video that he is still somewhat unstable but that he is able to do the essential mechanics of walking.
The thing is, he did not learn how to do this in a vaccuum. He had significant support from us, both in creating a safe environment for him to practice in (we added a rug to our living room and removed any sharp edges from his reach) and even helping him walk around by holding his hands. Most importantly, we modelled what walking looks like every day of his life so that he has a clear sense of the goal.
Walking is more efficient than crawling around, once you are proficient at it, but there is this awkward stage of learning how to walk when babies fall down a lot, cannot move very quickly, and when walking is actually not a very efficient mode of locomotion. If their parents did not walk, babies would not spontaneously learn how to do it on their own. They require models of what walking proficiently looks like so that they have a reason to continue to practice through the very inefficient early stages of learning how to walk.
Why then do we expect children to learn anything else when it is not modelled? Let us be clear here – telling children how something works is not modelling it. If we want children to learn how to think like mathematicians, for example, then we need to model how mathematicians think and act as mathematicians do.