A good discussion of the merits and perils of the Chinese script vs. alphabets.
So, this was the setting – small data sets, manual computation, and noisy environments. These were the conditions under which almost all the statistical procedures that we use today were produced. (p19)
A formula is a simple presciption for computation, one that does not contain data dependent branches. (p21)
I think what this allows us to do, and what is basically the trend for the future, is that we are substituting computer power for unverifiable assumptions about the data. […]
Why use these techniques? I think the reason is clear. The cost of computation is ever decreasing, but the price that we pay for the incorrect assumptions is staying the same. (p26)