Edge Cases Dominate the Problem

ML struggles with rare but critical scenarios; if edge cases matter most, explicit rules or expert systems may be better.