> They genuinely believed the current state of the art is polished and powerful enough to replace huge swathes of coders, designers, writers, etc. They already

> They genuinely believed the current state of the art is polished and powerful enough to replace huge swathes of coders, designers, writers, etc. They already have replaced designers and writers. It probably seems reasonable they could replace other jobs very soon, if your knowledge is superficial and you're just reading marketing brochures. The reality AI's are confidently wrong. Where that doesn't matter they already cut swathes through the workforce. Where it does they haven't made much progress so far. Yet, I saw a post from someone here saying they were a web programmer who didn't know javascript. They just prompted the AI until it produced something that worked. Them calling themselves a "web programmer" seemed like a stretch to me, and I'm sceptical about them tackling anything but cookie cutter work the AI has already seen a lot of, but that's clearly the direction copilot and it's ilk are headed. Currently, they look like they might soon succeed at using RAG to produce a better grep, so they have a little ways to go yet. I find checking the code written AI's to be more effort than it's work primarily because if I have to understand the code anyway, it's easier if I write it. But maybe that will change. One day. Right now this constant checking of AI's output is a huge time sink. But there is class of jobs that continually have humans checking their output and giving corrections. That is ... managers. Even CEO's. Maybe the future jobs of board is to select the right AI to be the CEO, and the future of the human workforce is to be prompt engineers for that AI CEO. My guess is AI's will be replacing managers and CEO's before they make a serious dent on STEM jobs.