So you're saying Goertzel believes

Just an off-the-cuff not-very-detailed hypothesis about what he believes.

with results equivalent for any reasonable initial mind design

Or at least any mind design that looks even vaguely person-like, e.g. uses clever Bayesian machine learning algorithms found by computational cognitive scientists; but I think Ben might be unknowingly ignoring certain architectures that are "reasonable" in a certain sense but do not look vaguely person-like.

And the only evidence for this is that it happened once.

Yes, but an embarrassingly naive application of Laplace's rule gives us a two-thirds probability it'll happen again.

This does look a little obviously epistemically unsound.

Eh, it looks pretty pragmatically incautious, but if you're forced to give a point estimate then it seems epistemicly justifiable. If it was taken to imply strong confidence then that would indeed be unsound.

(By the way, we seem to disagree re "epistemicly" versus "epistemically"; is "-icly" a rare or incorrect construction?)