This is a more extreme take than I would make, but I largely agree. I think the real-world capabilities of current AI are vastly overestimated. Are there areas

This is a more extreme take than I would make, but I largely agree. I think the real-world capabilities of current AI are vastly overestimated. Are there areas where it will create new possibilities? Absolutely. Will it make some workflows faster? Yeah, probably. Will it make software development faster? Maybe the really common stuff that "generic". I think in reality we'll see small marginal gains in a slightly better autocorrect for all but the most basic of CRUD and web dev. I just haven't seen the level of correctness and higher order understanding that's necessary for real software engineering. I keep seeing people saying they use it to generate tests, but are they also using it to generate the code under test? I'd be highly concerned with people relying on a tool that is inherently unreliable to do the majority of work. Human brains are bad interpreters, but AI is an even worse one. Maybe it will get there eventually, but I don't think GPT3/4 is it.