carboniferous_umbraculum comments on EIS V: Blind Spots In AI Safety Interpretability Research