Different kinds of super 'machine' intelligence are likely to exist in the future, with machines that are self improving. This is likely to provide a real challenge for human understanding. So constructing machines with values and deciding what those values should be is a huge challenge to philosophers, as Anders Sandberg explains.