People talk about the effectiveness (or lack thereof) of large language models as though all tasks are comparable. But it strikes me that there are three broad categories of work that large language models are applied to:

Compression.
Transformation.
Expansion.

Go to Source