I am so glad that this comment is at the top! Some other points - std.dev shows up in nature and is easier to understand, and analyze, simply because variances

I am so glad that this comment is at the top! Some other points - std.dev shows up in nature and is easier to understand, and analyze, simply because variances from independent sources add up. That's why something called ANOVA exists. In OLS regression, we minimize the std. dev. of the error, not the MAD. (It is not because statisticians "thought" std. dev is actually MAD, it's because minimizing std. dev or square error has very nice properties that we use it even though the solution to MAD minimization is also known.) Lastly, I don't follow who he proposes to be using MAD - if he decrees statisticians and physicists should still use Std. Dev? Often it is that I find these articles, and I let them pass by. But sometimes, they have a potential to come back - and I stand a chance to be quoted this article, by someone else, who I then have to drain my energy to make the point that it's not really worth reporting MAD from now on.