Brian Christian talks about the alignment problem in AI that makes it difficult to ensure that machines will be aligned with what we want them to do, and the terrifying implications of not fixing this issue.