It is really amazing how stubborn the DL community is in insisting on trying to chase infinity - i.e., trying to memorize the infinite object we call language. Data-driven/ML is a futile effort - we cannot model the innate knowledge structures that we employ in the process of language understanding by processing and seeing trillions and trillions and trillions of text patterns, because trillion divided by infinity is, guess what, ZERO!

https://medium.com/ontologik/memorizing-vs-understanding-read-data-vs-knowledge-d27c5c756740