My understanding of the original point is that ML models are trained on a data set. If you train the model on a data set that is only made up of white faces, th

My understanding of the original point is that ML models are trained on a data set. If you train the model on a data set that is only made up of white faces, the results will be whiter than if you train it on a data set that is made up solely of black faces. They said they'd like to know the makeup of the data set that the model is trained on to more fully understand how the results were produced.