Have you read a single paper in ai safety?