The Future of NLP

Vilia: I found one paper and the only, I think really relevant paper before my work in deep learning for NLP. But then none of their stuff worked, right? It's all just like, you start with pixels, they made sense,. They give you some signal and words are like, well, how do you map the word site kind of thought about it for a while. Vilia: Instead of recurring neural nets, we started to use recurrent neural nets which aren't as nicely parallelizable. And that led me to make them more palatable and viable for linguists  and natural language processing community.