Gary Marcus

THIS THIS THIS: “OpenAI has made it abundantly clear we can't trust them to disclose safety incidents.” (And see my substack this morning on the same point; wish I had this table!)