So basically if I ignored all impacts of using it I would be ok with it as a first pass at some sort of information gathering if I was willing to check deeper.
When I used it, I asked things like “change the tone of this writing” or “make these bullet points include key words to match this job post”.
That said, it does have impacts and I’m not willing to incur them for my passing curiosities or to speed up job applications.
One of my hobbies seems like it should be be easy for AI to move into. Crochet patterns are often fairly mathematically based. However every generated pattern I’ve looked at has been nonsensical garbage. Images that look cute and plausible at first pass are basically imagination. It used to be really easy to spot but it is getting harder/more subtle.
I used to work in healthcare and when I first got access to chat GPT I asked it some simple differential diagnosis questions. It gave a few very common possibilities for the symptoms I gave but completely ignore an important red flag that even a first year med student should (probably would) keep in mind.
There’s a bunch of people who are weirdly fascinated by a particular piece of hold music. I think Kaiser used to use it. NPR searched out the composer and I have a file of the whole thing.
Very rarely and usually only in cases where it is either exceptional or awful.