[lunias]

I'm not sure the people writing these articles actually use AI outside of trivial benchmarks. Sure, it can create "a website" in minutes, but what does that website do? I've found (I'm working on vehicle routing and maintenance scheduling) that you very much have to have a plan in your head and an understanding of what's going on in order to build towards reasonable results effectively. There is no replacement for knowing what you want, how to ask for it, and how to verify the implementation. That takes expertise. You cannot tell AI, "We fired the engineers. Take over their jobs and do them better." That's a non-starter.

[kjellsbells]

Correctness like this doesn't really matter, if you are some EVP at BigCorp: you don´t really need AI to be genuinely superior to your human workers. You need AI to appear to be useful enough for two or three quarters, max. It gives you cover to justify eliminating jobs and still have the stock price rise. It gives you exposure to the people who might hire you for your next golden gig ("thought leader", "visionary" etc). And you'll be bailing shortly anyway, before reality comes knocking.

[kj4211cash]

It sounds like you are an Operations Researcher. I am too friend. The talks I've seen on people using AI for Operations Research have been ... underwhelming. I wonder if the vast majority of us are just in niches narrow enough that AI just isn't that helpful.

[lunias]

Lack of training data in narrow domains is definitely true. I'm hopeful though, AI seems like it should mesh well with the problem. One of the biggest wins for me so far is just how quickly I can build and iterate on visualizations of the data. It has cut down enormously on the amount of time that I'm asking myself, "but why is this the schedule?".