The problem for them is that to make a Large Language Model, is it needs to consume a large amount of language. Meaning, to get even a semblance of basic functionality - literally putting together a grammatically correct sentence - means it will need to consume many orders of magnitude more language examples than their right-wing bullshit mill has ever produced. It could consume the entirety of Fox News’ catalogue and not reach even a fraction of what most LLM models have been trained on.

It’s currently impossible for an LLM to exist without having been trained on an absolute assload of progressive literature.