Of course it can! You can train any bias you want into your neural net. The question is how long your engineers will buy it, and whether the socially acceptable default position, that men and women should expect identical outcomes given identical opportunities, is actually true - because that is neither theoretically not empiricially verified, despite what outlets like NYT [and third wave feminists in general] are desperate to have you believe.
Before you flag and downvote, please seriously consider my last point, a reckoning may be quickly approaching Western society. We have taken for granted the conflation of equality of opportunity with equality of outcome - how will Western society respond if (I'm not saying when, but if) neural nets consistently predict gendered differences? Will we revert some of our recent assumptions regarding human nature, or will we collectively insist that the data is bad until it is massaged to show the results we expect?
1 comment
[ 3.2 ms ] story [ 10.6 ms ] threadBefore you flag and downvote, please seriously consider my last point, a reckoning may be quickly approaching Western society. We have taken for granted the conflation of equality of opportunity with equality of outcome - how will Western society respond if (I'm not saying when, but if) neural nets consistently predict gendered differences? Will we revert some of our recent assumptions regarding human nature, or will we collectively insist that the data is bad until it is massaged to show the results we expect?