Logan Zoellner comments on an effective ai safety initiative