(June 30 23:52) Forbes.com
'It is not that you can train an algorithm and and it will warn you that its doing something that may be wrong. Whatever bias was in that training set will come at you full force when that algorithm runs.'
You can find the original article
here