Sam Altman’s Chip Ambitions Undercut OpenAI’s Safety Strategy — LessWrong