The Legitimacy of the Unintelligible Physics

The deep learning revolution cultivated the emergence of predictive tools, known as oracles, that do not directly improve understanding of predicted phenomena. The rise of oracles conjures an alarming vision of a new physics devoid of understanding. This unsettling prospect raises the question of whether models that offer little understanding can nevertheless be legitimate scientific tools. We examine the relationship between oracles and conventional physical theories through their mathematical structure, ontological variables, and empirically determined constants. We propose a comparative metric, I , based on the ratio of empirically determined constants to ontological variables. Under consistently applied counting and representation conventions, I characterizes one proposed source of epistemic opacity; a high value of I does not by itself establish intelligibility. This approach offers a framework for investigating differences in intelligibility between oracles and conventional physical theories without treating them as categorically separate forms of scientific inquiry. We argue that limited intelligibility does not, by itself, disqualify empirically validated oracles as legitimate tools for use in exploratory science, where they can support prediction, reveal patterns, and guide discovery.