Ebenezer Dukakis comments on Can we safely automate alignment research?