What is Large Language Models?

Large Language Models are neural models trained on large text corpora to predict and generate language. In practice, the value of this concept comes from understanding where it belongs in the architecture, what problem it solves, and what constraints or tradeoffs it introduces.