On Sat, 20 Mar 2021 11:11:03 -0400 Kevin A. McGrail wrote:
> Would be worth looking at how it got through. Perhaps badly trained > Bayesian learning, for example? If it's done well it can look like a hybrid of a real paypal email and a routine gmail email and Bayes isn't capable of spotting the incongruity. It isn't necessarily a sign of mistraining if some of these hit BAYES_00.