Great discussion Mike. New model of science indeed. But much of AI relies on Bayesian Inference rather than strict causality. In certain situations it becomes a model science without models. Now that we know the systems can "hallucinate", we might go off half-cocked on spurious interconnections and waste time on half-baked understandings. And the deeper deep learning (layered not profundity) goes, the more difficult it is to document exactly how one got there. But it is exciting.