Logic of Hypotheses: from Zero to Full Knowledge in Neurosymbolic Integration