On Automation and LLMs

A case study for how I use LLMs in my day to day work Let me preface this by saying I’m in general very resistant to change. I like to wait out a little bit for the hype to settle and have a better understanding of how things work before jumping head first. What LLMs are good at? Risking repeating things that other people said better 1 , let me try and give my view on what LLMs are good at. It has been said that LLMs are “auto complete on steroids”, this is a decent analogy, but only partially true. Auto complete is rudimentary in the sense that it only takes into account a few previous words that you typed to predict the next one. The thing that makes LLMs shine is it models correlations between words, and to an extent 2 , concepts, in measurable mathematical terms. So for example, we…