Joint probability goes both ways: Ng's prediction that X% of workers will become obsolete is just as unlikely as any future, including Bostroms, as any long string of probable outcomes is highly unlikely.