Donald Hobson answers Is AI safety doomed in the long term?