I guess, in a sense, the intelligence "black box" up to now, in science, has been the <i>human</i> mind ...<p>... and we built mechanisms in the scientific process to provide visibility into what are human, internal proceses (peer review, references, data openness and others) ...<p>... we need to to likewise with AI.-<p>(Not that it's easy. But it seems to me like an analogous scenario ...)