Christopher Olah comments on Anthropic’s Core Views on AI Safety