It seems kind of silly to draw broad conclusions about the capabilities of text generation in a specific domain using a general purpose language model. Minimall

It seems kind of silly to draw broad conclusions about the capabilities of text generation in a specific domain using a general purpose language model. Minimally, they should have fine-tuned it on law journal articles.