[devjab]

I genuinely find ChatGPT very useful for a lot of tedious tasks, but it's also very obvious that it's data cutoff point was in 2021. You can get it to write you an abstract generic Odata controller class in C# better than most searching on google, but it'll be with IActionResult and not ActionResult, sometimes it'll even be with Task<Iqueryable>, which is even older and it'll do this extremely confidently. Which is sort of fine, it'll work after all. It's also not fine, because there is a reason it's moved on to become ActionResult performance and ease of maintenance being some of the reasons.

Though to be fair, it's not like Stack Overflow, or even some of the less updated parts of the official documentation, will do any better than ChatGPT, but with ChatGPT you don't get a date on the knowledge and you don't get the comments from other people telling you that it's wrong or outdated. For me personally, there is also some thing about its confidence that makes me "trust" it more than the internet that I've spent 20+ years not trusting. I'm not sure if that's just me or even why that is exactly. I'm fully aware that the language model is basically just the internet, and still I believe it? I'm happy the first thing I asked it was on a subject I knew a lot about so that I could see straight away that its answer was very outdated. Because if it had been on a subject I didn't know much about, I'm not sure I would have even found out its cutoff point was in 2021. I only learned that fact because the answer it suggested was with a library that I knew was abandoned, to which I asked if it knew that, and it told me when it had stopped "learning".