I've not found them reliably useful in a personal or professional capacity for anything other than trivial tasks. I've spent a lot of time reviewing code written by LLMs and it was not a good use of my time.

A technology that is inherently unreliable and requires constant supervision needs to be on the level of nuclear fission to be worth the hassle, in my view.

Note that I specifically am talking about "hoover up the internet" LLMs, not specialist tools built on scientific datasets or whatever.

It can summarise my email? Fantastic, except it cannot be relied on to do a good job so I might as well do the reading myself as at least that way the data I'm getting is correct.