How is it possible to "generalize" from one data point? The normal method of generalizing is finding a similarity shared across many data points, and proposing it applies more widely.

Perhaps one could generalize badly with 2 data points, but isn't it just impossible with 1, and therefore he must actually be doing something else?