Raw LLM Responses

I never understood this. My best guess is that programmers are on the forefront of LLM's capabilities and have integrated it into their workflows probably way faster than any other discipline. People in, or adjacent to the field, surprised by it's effectiveness just jumped the gun on overstating it's capabilities. What I find funny is they are called large LANGUAGE models, if anyone is getting replaced first it's gana be lawyers or the people who do the grunt work for them (para legals? i dunno). The best use i've found for AI seems like it would be perfect for lawyers or really anyone who has to do lots of research across many different documents. Imagine you have a case and you can literally ask an LLM that only has access to real cases in it's database and that's setup to always return links to the hard references it's pulling information from. I feel like that contextual search power combined with restrictive datasets, it's summarizing ability and the ability to have it always leave real references for double checking would be huge. We've been using LLM's to do just that like, months after chatGPT first blew up.