Comment by will-burner

> It's rare to see NLP research that doesn't have a dependency on closed data controlled by OpenAI and Google, two companies that I already despise.

The dependency on closed data combined with the cost of compute to do anything interesting with LLMs has made individual contributions to NLP research extremely difficult if one is not associated with a very large tech company. It's super unfortunate, makes the subject area much less approachable, and makes the people doing research in the subject area much more homogeneous.