Up in arms for what reason? That neural networks are perfectly interpretable? That's the nature of huge amorphous deep networks. These feature extraction forays are a good step forward though.