Comment by cold_harbor

worth separating: LSTM (Hochreiter & Schmidhuber 1997) is ironclad and widely cited. the transformer attention priority claims are far shakier. conflating them is how Schmidhuber undermines himself