Thus when unsure we do not simply go with the best assumption.
"AI"s seems to just go with whatever is the "most" likely, even if it is just some 0.0001% above a sea of alternatives.