This blog discusses research ideas about building AI safely and responsibly from the Google DeepMind alignment team.