AI alignment to human values is surprisingly hard task. And extraordinary important one, as unaligned AI is extinction level risk (not only by the Skynet scenario, but even much subtle ones).