quetzal_rainbow comments on Interpretability Will Not Reliably Find Deceptive AI