I was prompted here by reading this excellent paper by @MelMitchell1 & David Krakauer - & now find this in it: "can [LLMs'] unimaginably large systems of statistical correlations produce abilities that are functionally equivalent to human understanding?"