The alignment problem from a deep learning perspective https://arxiv.org/abs/2209.00626 Richard Ngo, Lawrence Chan, Sören Mindermann Alignment problem in AI