I concur with both. I add the caveat that the 'taking ideas seriously' rational vampire would clearly be best served by vamping willing FAI researchers. That could be expected to raise p(FAI | GAI) up to well above real world values. I say above just because it eliminates some significant contributors towards error (memory failure, fatigue, time pressure and cognitive decline from aging).