[YeGoblynQueenne]

Thanks, but I wanted to hear why the OP thinks LSTM gates are elegant and add to interpretability.

The OP made a big todo about how the article above is misguided and the person who wrote it has a poor understanding of deep learning, so I'm assuming he or she has a very good understanding of the subject.

I mean, I'd hate to believe the OP is just playing deep learning bingo for HN karma, you know?