How can we adequately use AI when there is inherent bias built into these engines?