ETHICS in Machine Learning

Talk about how to make Machine Learning ethical and potential ethical concerns.

Once got asked by a government customer if we could track people in their airports ‘by skin color’.

Why was accuracy for black women so low? Just Lack of training data? Does architecture need to be tailored to handle different ethnicity? Could
Loss function treat Minority faces as more important to offset for less examples in training data and therefore less total impact on loss?

1 Like