The biggest difference to me is that we have humans that claim they can explain why compilers work the way they do. But I might as well trust someone who says the same about LLMs, because honestly I have no way to verify if they speak the truth. So I am already offloading a lot of burden of proof about the systems I work on to others. And why does this ”other” need to be a human.