faul_sname comments on Interpretability Will Not Reliably Find Deceptive AI