I don't think it's just LLMs, but I do think it's reasonable to think of it as automating data-science. I mean that to describe how I see its scope of applications, rather than a bah humbug, and of course the scope of data-science is huge. Being able to automate specific analyses sounds like game changer

That's all sort of esoteric though - to most people, AI is being deliberately presented over and over in human-intelligence terms, and that's how they learn to understand it. Like, none of my coworkers approach me to ask what I think about AI's potential uses in chemistry, content moderation, copywriting, statistical modeling, catching relations between data that people aren't very well suited to catching, or anything else plausible. They want to know if their geekest coworker is worried about The Singularity

So for the general public, who unironically seem to think of AI in a "Skynet/not Skynet" dichotomy, I tell them "not Skynet"