davekasten comments on Provably Safe AI: Worldview and Projects