If "superintelligent AI" will learn from us rather than making up it's own values (which seems likely to me) then we should be a good example to it.