so, e.g., in ML/AI in 2026, the subsymbolic notion of learning is so contextually salient that denying some capacity is learned/learnable can often reasonably be read as incompatible with induction by a domain-general NN; in the case of fodor's argument about learnability, it should not be.
